<?xml version="1.0" encoding="utf-8"?>
  <?xml-stylesheet type="text/xsl" href="rfc2629.xslt" ?>
  <!-- generated by https://github.com/cabo/kramdown-rfc version 1.6.17 (Ruby 3.0.2) -->


<!DOCTYPE rfc  [
  <!ENTITY nbsp    "&#160;">
  <!ENTITY zwsp   "&#8203;">
  <!ENTITY nbhy   "&#8209;">
  <!ENTITY wj     "&#8288;">

<!ENTITY RFC7011 SYSTEM "https://bib.ietf.org/public/rfc/bibxml/reference.RFC.7011.xml">
<!ENTITY RFC8986 SYSTEM "https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8986.xml">
<!ENTITY RFC9315 SYSTEM "https://bib.ietf.org/public/rfc/bibxml/reference.RFC.9315.xml">
]>


<rfc ipr="trust200902" docName="draft-irtf-nmrg-ai-challenges-07" category="info" submissionType="IRTF" tocInclude="true" sortRefs="true" symRefs="true">
  <front>
    <title abbrev="Coupling AI and network management">Research Challenges in Coupling Artificial Intelligence and Network Management</title>

    <author fullname="Jérôme François">
      <organization>University of Luxembourg and Inria</organization>
      <address>
        <postal>
          <street>6 Rue Richard Coudenhove-Kalergi</street>
          <city>Luxembourg</city>
          <country>Luxembourg</country>
        </postal>
        <email>jerome.francois@uni.lu</email>
      </address>
    </author>
    <author fullname="Alexander Clemm">
      <organization>Sympotech</organization>
      <address>
        <postal>
          <city>Los Gatos, CA</city>
          <country>United States of America</country>
        </postal>
        <email>ludwig@clemm.org</email>
      </address>
    </author>
    <author fullname="Dimitri Papadimitriou">
      <organization>Université Libre de Bruxelles (ULB), and He2B - ESI</organization>
      <address>
        <postal>
          <street>Boulevard du Triomphe - CP 212</street>
          <city>Brussels</city>
          <country>Belgium</country>
        </postal>
        <email>papadimitriou.dimitri.be@gmail.com</email>
      </address>
    </author>
    <author fullname="Stenio Fernandes">
      <organization>Canada Post</organization>
      <address>
        <postal>
          <street>2701 Riverside Drive</street>
          <city>Ottawa, ON K1A 0B1</city>
          <country>Canada</country>
        </postal>
        <email>stenio.fernandes@ieee.org</email>
      </address>
    </author>
    <author fullname="Stefan Schneider">
      <organization>Digital Railway (DSD) at Deutsche Bahn</organization>
      <address>
        <postal>
          <country>Germany</country>
        </postal>
        <email>stefanbschneider@outlook.com</email>
      </address>
    </author>

    <date year="2026" month="November" day="20"/>

    
    <workgroup>Internet Research Task Force</workgroup>
    <keyword>network management</keyword> <keyword>artificial intelligence</keyword> <keyword>machine learning</keyword>

    <abstract>


<t>This document is intended to introduce the challenges to overcome when Network Management (NM) problems may require coupling with Artificial Intelligence (AI) solutions. On the one hand, many difficult NM problems still lack good solutions, or existing approaches come with significant limitations. Artificial Intelligence may help produce novel solutions to those problems. On the other hand, due to the high computational costs of AI solutions and stringent data privacy constraints, the distributed execution of AI workloads has become paramount. Consequently, networks must be operated efficiently to sustain these distributed processing requirements.</t>

<t>To identify the right set of challenges, the document defines a method based on the evolution and nature of NM problems. This will be done in parallel with advances and the nature of existing solutions in AI in order to highlight where AI and NM have already been coupled together or could benefit from a closer integration. So, the method aims at evaluating the gap between NM problems and AI solutions. Challenges are derived accordingly, assuming that solving these challenges will help to reduce the gap between NM and AI.</t>

<t>This document is a product of the Network Management Research Group (NMRG) of the Internet Research Task Force (IRTF).  This document reflects the consensus of the research group.  It is not a candidate for any level of Internet Standard and is published for informational purposes.</t>



    </abstract>



  </front>

  <middle>


<section anchor="intro"><name>Introduction</name>

<t>The functional scope of Network Management (NM) is very large, ranging from monitoring to accounting, from network provisioning to service diagnostics, from usage accounting to security. The taxonomy defined in <xref target="Hoo18"/> extends the traditional Fault, Configuration, Accounting, Performance, Security (FCAPS) domains by considering additional functional areas but above all by promoting additional views. For instance, network management approaches can be classified according to the technologies, methods or paradigms they will rely on. Methods include common approaches as for example mathematical optimization or queuing theory but also techniques which have been widely applied in last decades like game theory, data analysis, data mining and machine learning. In management paradigms, autonomic and cognitive management are listed. As highlighted by this taxonomy, the definition of automated and more intelligent techniques has been promoted to support efficient network management operations.</t>

<t>Research in NM and networking has been very active in applied ML <xref target="Bou18"/> and has enabled progress in areas such as network security <xref target="Ke23"/>, Software-Defined Networks (SDN) <xref target="Xie19"/>, and vehicular networks <xref target="Tan21"/>. This broad adoption has led to a strong interlinkage between networking and AI, laying part of the foundation for future systems such as 6G or B5G networks that are expected to be AI-native <xref target="Sha25"/>.</t>

<t>However, for maintaining network operational in pre-defined safety bounds, NM still heavily relies on established procedures. Even after several cycles of adding automation, these procedures are still mostly fixed and set offline in the sense that the exact control loops and all possible scenarios are defined in advance. They are so mostly deterministic by nature or at least with sufficient safety margin. Obviously, there have been a lot of propositions to make network smarter or more intelligent with the use of Machine Learning (ML) but without large adoption for running real networks because it changes the paradigms towards stochastic methods.</t>

<t>ML includes regression analysis, statistical learning (SVM and variants), deep learning (ANN and variants), Reinforcement Learning (RL), Large Language Models (LLMs), etc. It is a sub-area of Artificial Intelligence (AI) that concentrates the focus nowadays but AI encompasses other areas including knowledge representation, inductive logic programming, inference rule engine or by extension the techniques that allow one to observe and perform actions on a system.</t>

<t>It is thus legitimate to question whether ML or AI in general could be helpful for NM in practical deployment. This question is closely tied to the problems NM aims to address. Independently of NM, ML-based solutions were introduced to solve problems in an approximate way when they are very complex in nature, i.e., when finding an optimal solution is not possible in polynomial time. This is the case for NP-hard problems. In those cases, solutions typically rely on heuristics that may not yield optimal results, or algorithms that run into issues with scalability and the ability to produce timely results due to the exponential search space. In NM, those problems exist: allocation of resources in case of Service Function Chaining (SFC) or network slicing among others are recent examples which have gained interest in our community with SDN. Many proposed approaches consist of modelling the optimization problem as a MILP (Mixed-Integer Linear Programming) and solve it by the means of heuristics to reach a satisfactory trade-off between solution quality (gap to the optimal solution) - computation time and model size/dimensionality. Hence, ML is recognized to be well adapted to progress on this type of problem <xref target="Kaf19"/>.</t>

<t>However, not all computational problems in NM are NP-hard. Due to real-time constraints, some involve very short control loops that require both rapid decisions and the ability to quickly adapt to new situations and different contexts. In such cases, time is critical, and approximate solutions are often acceptable. This is another domain where AI can be beneficial. Expert systems are AI systems <xref target="Ste92"/>, but rule-based systems are not designed to scale with the volume and heterogeneity of data that can be collected in a network today. In contrast, ML is better suited to automatically learn abstract rule structures that can be updated over time.</t>

<t>Another common problem in NM is classification. For instance, classifying network flows is helpful for security purposes to detect attack flows and to differentiate QoS among flows (e.g., real-time streams that require prioritization). ML-based classification algorithms have been widely used with satisfactory results when properly applied and have been adopted in commercial products. There are many algorithms including decision trees, Support Vector Machine (SVM) or (deep) Neural Networks (NNs) which have been to be proven to be effective in many areas and notably for image and natural language processing.</t>

<t>Finally, many problems still involve humans in the loop: from support issues such as handling trouble tickets to planning activities for the rollout of new services. This creates operational bottlenecks and is often expensive and error-prone. These tasks could be either automated or guided by an AI system to avoid individual human bias. It is worth noting that ML models trained on human-generated data can also indirectly suffer from collective human bias. Indeed, the balance between human resources and problem complexity is increasingly skewed and will continue to worsen due to growing network size, device heterogeneity, service diversity, and more. As a result, human-based procedures tend to be either overly simplistic for the problems they address or excessively time-consuming. Notable examples are in security, where network operators must defend against potential unknown threats.</t>

<t>All these problems are exacerbated by increasingly complex networks operating across many dimensions (users, devices, services, connections, etc.). Therefore, AI is expected to enable or simplify solving these problems in real networks in the near future <xref target="czb20"/> <xref target="Yan20"/>, as they require unprecedented levels of performance in terms of throughput, latency, mobility, security, and more.</t>

</section>
<section anchor="acronyms"><name>Acronyms</name>

<t><list style="symbols">
  <t>AI: Artificial Intelligence</t>
  <t>CNN: Convolutional Neural Network</t>
  <t>FL: Federated Learning</t>
  <t>GAN: Generative Adversarial Network</t>
  <t>GNN: Graph Neural Network</t>
  <t>IBN: Intent-Based Networking</t>
  <t>LLM: Large Language Model</t>
  <t>LSTM: Long Short-Term Memory</t>
  <t>MAE: Mean Absolute Error</t>
  <t>ML: Machine Learning</t>
  <t>MILP: Mixed-Integer Linear Programming</t>
  <t>MLP: Multilayer Perceptron</t>
  <t>MSE: Mean Squared Error</t>
  <t>NM: Network Management</t>
  <t>RL: Reinforcement Learning</t>
  <t>SDN: Software-Defined Networks</t>
  <t>SFC: Service Function Chaining</t>
  <t>SVM: Support Vector Machine</t>
  <t>VNF: Virtual Network Function</t>
</list></t>

</section>
<section anchor="pbNM"><name>Difficult problems in network management</name>

<t>As mentioned in the introduction, problems to be tackled in NM tend to be complex and exhibit characteristics that make them good candidates for AI-based solutions. We summarize these characteristics with short names to improve readability and to help the reader track how they recur throughout the document:</t>

<t><list style="symbols">
  <t>C1: Fundamental Algorithmic Complexity -- Managing a very large solution space that combinatorially explodes with the scale and topological footprint of the network domain, making it computationally impractical to explore and evaluate every possible state space (such as in NP-hard problems).</t>
  <t>C2: Uncertainty and Unpredictability -- Coping with highly dynamic, non-deterministic variables along multiple operational dimensions, including shifting user traffic patterns, lack of visibility into end-to-end network states, and myriads of multi-domain provider interdependencies that cannot be modeled or known in advance.</t>
  <t>C3: Real-Time Constraints and Determinism -- Computing analytical solutions, issuing telemetry verdicts, and executing closed-loop control actions within strictly bounded or deterministic timeframes, especially where the context changes dynamically and decisions must be implemented rapidly to remain useful.</t>
  <t>C4: Data Dependency and Heterogeneity -- Processing and analyzing massive volumes of network telemetry to derive accurate management decisions, which requires dealing with compound data challenges ranging from data heterogeneity and incomplete or missing datasets to high data velocity.</t>
  <t>C5: Integration with Existing Practice -- Interoperating seamlessly with pre-existing automated procedures, legacy deterministic control loops, heterogeneous operator workflows, and established human-in-the-loop administrative structures.</t>
  <t>C6: Economics and Computational Footprint -- Ensuring cost-effectiveness across limited physical and operational resources (including available link bandwidth, node CPU, and human overhead), particularly when model training or inference processing is distributed at the network edge or embedded within data-plane elements.</t>
</list></t>

<texttable title="Problem characteristics" anchor="char-table">
      <ttcol align='left'>Characteristic Tag &amp; Short Name</ttcol>
      <ttcol align='left'>Relevant Document Sections / Gaps</ttcol>
      <c>C1: Fundamental Algorithmic Complexity</c>
      <c><xref target="problemMapping"></xref> (Problem Mapping); <xref target="planningActions"></xref> (Action Planning); <xref target="DistributedAI"></xref> (Distributed AI)</c>
      <c>C2: Uncertainty and Unpredictability</c>
      <c><xref target="planningActions"></xref> (Action Planning); <xref target="performanceModels"></xref> (Performance of produced models); <xref target="securityOfAI"></xref> (Security of AI); <xref target="relevanceOfAIOutput"></xref> (Relevance of AI-based outputs)</c>
      <c>C3: Real-Time Constraints and Determinism</c>
      <c><xref target="lightweightAI"></xref> (Lightweight AI); <xref target="productionSystems"></xref> (Production Systems); <xref target="performanceModels"></xref> (Performance of produced models)</c>
      <c>C4: Data Dependency and Heterogeneity</c>
      <c><xref target="dataForAI"></xref> (Network data as input for ML algorithms); <xref target="dataCollection"></xref> (Data collection); <xref target="usableData"></xref> (Usable data)</c>
      <c>C5: Integration with Existing Practice</c>
      <c><xref target="humansInLoop"></xref> (AI with humans in the loop); <xref target="netManFordistributedAI"></xref> (NM for AI); <xref target="explainability"></xref> (Explainability)</c>
      <c>C6: Economics and Computational Footprint</c>
      <c><xref target="lightweightAI"></xref> (Lightweight AI); <xref target="productionSystems"></xref> (Production Systems)</c>
</texttable>

<t>Many decision/optimization problems are affected by multiple criteria. Below is a non-exhaustive list of complex NM problems for which AI and/or non-AI-based approaches have been proposed:</t>

<t><list style="symbols">
  <t>Computation of optimal paths: Packet forwarding is not always based on traditional least-cost routing protocols, but on computation of paths that are optimized for certain criteria - for example, to meet certain level objectives, to result in greater resilience to balance utilization, to optimize energy usage, etc.  Many of those solutions can be found in SDN, where a controller or path computation element computes paths that are subsequently provisioned across the network.  However, such solutions generally do not scale to millions of paths (C1) and cannot be recomputed in sub-second time scales (C3) to take into account dynamically changing network conditions (C2). To compute those paths, operations research techniques have been extensively used in literature along with AI methods <xref target="Lop20"/>. Mobility and dynamicity are two conditions that make the problem of computing optimal paths even harder. Hence, adaptive routing based on RL has been proposed as it allows an agent to promptly react to topological changes <xref target="Sin22"/>. As such, this problem can be considered as close to big data problems with some of the different Vs: volume, velocity, variety, value…</t>
  <t>Classification of network traffic: Without loss of generality, a common objective of network monitoring for operators is to know the type of traffic going through their networks (web, streaming, gaming, VoIP). Such a task analyses data (C4) which can vary over time (C2) except in very particular scenarios like industrial isolated networks. However, the output of the classification technique is time-constrained only in specific cases where fast decisions must be made, for example to reroute traffic. Simple identification based on IANA-assigned TCP/UDP port numbers were sufficient in the past. However, with applications using dynamic port numbers, signature techniques can be used to match packet payload <xref target="Sen04"/>. To handle applications now encapsulated in encrypted web or VPN traffic, various ML-based approaches have been thus adopted <xref target="Bri19"/><xref target="Naj24"/><xref target="Wan24"/><xref target="Ake24"/> including LLMs <xref target="Gin24"/>.</t>
  <t>Network diagnostics: Disruptions of networking services can have many causes and thus can rely on analysing many sources of data (C4). Identifying the root causes is of high importance, so that repair actions can address them versus just working around the symptoms. Such repair actions may involve human actions (C5). Further complicating the matter are scenarios in which disruptions are not complete but involve only a degradation of service level, and where disruptions are intermittent, not reproducible, and hard to predict.  Artificial intelligence techniques can offer promising solutions. Especially, anticipation of faults is of paramount importance and will lead to the development of predictive maintenance in future networks<xref target="Mut24"/>.</t>
  <t>Network observability: obtaining deeper insight into network status can rely on monitoring techniques to gather data from various sources. A major issue is to aggregate all these data in a valuable format <xref target="Zha21"/>. When it is not directly used to automate some actions, the aggregation of the data needs to be presented in an interpretable manner to human operators. In this area, visualisation techniques are helpful and also rely on AI techniques to provide the best outputs by reducing the number of dimensions (C4) and adapting the visualisation of data for human-operated processes (C5) <xref target="Ami24"/>.</t>
  <t>Intent-Based Networking (IBN): Roughly speaking, IBN refers to the ability to manage networks by articulating desired outcomes without the need to specify a course of action <xref target="RFC9315"/>.  The ability to determine such courses of actions, in particular with multiple interdependencies, conflicting goals, large scale, and highly complex and dynamic environments is a huge and largely unsolved challenge (C1, C2, C3). As an illustration, a major problem with intent is to interpret them correctly knowing that different intent formats have been proposed including natural language. Without good interpretation of the intent, i.e. the expected outcomes, the derived actions will not be adequate. In case the intent is correctly interpreted, a major problem is to find concrete solutions to realize the intents which implicitly requires optimizing the actions to be taken. Artificial Intelligence techniques can be of help here in multiple ways, from accurately classifying dynamic context to determine matching actions to reframing the expression of intent as a game that can be played (and won) using artificially intelligent techniques.  As an example, LossLeaP even goes further by trying to predict the gap between a targeted objective and the predicted impact of the intent realization <xref target="Col22"/>.</t>
  <t>VNF (Virtual Network Function) placement and SFC (Service Function Chain) design: VNFs need to be placed on physical resources and SFCs designed in an optimized manner to minimize the use of networking resources and energy (C1,C6). As it is known to be a NP-hard problem, many heuristic- or machine-learning-based approaches have been proposed. The VNF paradigm actually emerges alongside 5G networking and orchestration methods <xref target="Att23"/>.</t>
  <t>Smart admission control to avoid congestion and oversubscription of network resources: Admission control needs to be set up to ensure service levels are optimized in a manner that is fair and aligned with application needs, congestion avoided, or its effects mitigated (C6). This field of research has notably been extended to the context of network slicing <xref target="Vin21"/><xref target="Sul23"/>.</t>
</list></t>

</section>
<section anchor="hlchallenges"><name>High-level challenges in adopting AI in NM</name>

<t>As shown in the previous section, AI techniques are good candidates for difficult NM problems. There have been many proposed solutions, but most remain at the prototype level or have been evaluated only through simulation and/or emulation. It is thus questionable why the NM community conducts extensive research in this direction but has not widely adopted those solutions to operate real networks. There are several obstacles.</t>

<t>First, AI advances have been historically driven by the image/video, natural language and signal processing communities as well as robotics for many decades. As a result, the most impressive applications are in this area including recently the generalization of home assistants, chatbots or the large progress in autonomous vehicles. However, the network experts have been focused on building the Internet, in particular designing protocols to make the world interconnected and with always better performance and services. This trend continues today with the 5G networks in deployment and beyond 5G under definition. Hence, AI was not the primary focus even if increased network automation calls for AI and ML solutions. However, AI is now considered as a core enabler for the future 6G networks which are sometimes qualified as AI-native networks <xref target="Sha25"/>.</t>

<t>While we can see major contributions in AI-based solutions for networking over more than two decades, only a fraction of the community was concerned by AI at that time. Progress as a whole, from a community perspective, was therefore limited and compensated by relying on the development of AI in the communities as mentioned earlier. Even if our problems share some commonalities, for example on the volume of data to analyse, there are differences: data types are completely different, networks are by nature heavily distributed, etc. If problems are different, they should require distinct solutions or at least in-depth adaptation. In a nutshell, network-tailored AI has been overlooked, leading to a first set of challenges described in <xref target="AItechniques"></xref>.</t>

<t>Second, many AI techniques require sufficiently representative data. For example, (deep) learning techniques mostly rely on having vectors of (real) numbers as input which fits naturally with numerical metrics (packet/byte counts, latency, delays, etc.) but needs some adjustment for categorical (IP addresses, port numbers, etc.) or topological features. Conversions are usually applied using common techniques like one-hot encoding or by coarse-grained representations <xref target="Sco11"/>. However, more advanced techniques can be proposed to embed representation of network entities rather than pure encoding as illustrated in <xref target="Rin17"/><xref target="Evr19"/><xref target="Sol20"/>.</t>

<t>Besides, AI techniques that involve analysis of networking data can also lead to the extraction of sensitive and personally identifiable information, raising potential privacy concerns and concerns regarding the potential for abuse. Actually, this is a common and known problem that applies to many application domains <xref target="Liu21b"/>. For example, AI techniques used to analyse encrypted network traffic with the legitimate goal to protect the network from intrusions and illegitimate attack traffic could be used to infer information about network usage and interactions of network users <xref target="Hoa21"/>.  Intelligent data analysis and the need to maintain privacy are in many ways contradictory in nature, resulting in an arms race.  Similarly, training ML solutions on real network data is often preferable over using less-realistic synthetic data sets <xref target="Liu22"/>. However, network data may contain private or sensitive data, the sharing of which may be problematic from a privacy standpoint and even result in legal exposure. The challenge concerns thus how to allow AI techniques to perform legitimate network management functions and provide network owners with operational insights into what is going on in their networks, while prohibiting their potential for abuse for other (illegitimate) purposes.  Challenges related to network data as input to ML algorithms is detailed in <xref target="netData"></xref>.</t>

<t>Finally, networks are already operated thanks to (semi-)automated procedures involving many resources which are synchronized with management or orchestration tools. Adding AI therefore requires seamless integration within pre-existing processes. Although the goal of these procedures might be solely to provide relevant information to operators through alerts or dashboards in case of monitoring applications, this can be defined to trigger actions on the different resources, which can be local or remote. The use of AI or any other approaches to derive NM actions adds further constraints on them, especially regarding time constraints <xref target="Liu21"/> and synchronization to maintain a coherence over a distributed system.</t>

<t>A related challenge concerns the fact that to be deployed, a solution needs to provide a technical solution but also be acceptable to users - in this case, network administrators and operators.  With automated solutions concerns that users want to feel “in control” and able to understand what is going on, even more so if ultimately those users are held accountable for whether or not the network is running smoothly.  To mitigate those concerns, aspects such as the ability to explain actions that are taken - or about to be taken - by AI systems become important <xref target="Sen24"/>.</t>

<t>Beyond reasons of making users more comfortable, there are potentially also legal or regulatory ramifications to ensure that actions taken are properly understood.  For example, agencies such as the FCC may impose fines on network operators when services such as E911 experience outages.  In investigating causes for such outages, the underlying behaviour of the systems has to be properly understood, and even more so the reasons for actions that fall under the realm of network operations. All these aspects about integration and acceptability of the integration of AI in NM processes is detailed in <xref target="acceptability"></xref>.</t>

</section>
<section anchor="AItechniques"><name>AI techniques and network management</name>

<section anchor="problemMapping"><name>Problem type and mapping</name>

<t>An increasing number of different AI techniques have been proposed and applied successfully to a growing variety of different problems in different domains, including network management <xref target="Bou18"/>. Some of the most recently proposed AI approaches are clearly advancements of older approaches, which they supersede. Many other AI approaches are not predecessors or successors but simply complementary because they are useful for different problems or optimize different metrics. In fact, different AI approaches are useful for different kinds of problem inputs (e.g., tabular data vs. text vs. images vs. time series) and also for different kinds of desired outputs (e.g., a predicted value, a classification, or an action). Similarly, there may be trade-offs between multiple approaches that take the same kind of inputs and desired outputs (e.g., in terms of desired objective, computation complexity, constraints).</t>

<t>Overall, it is a key challenge of using AI techniques for network management to properly understand and map which kind of problems with which inputs, outputs, and objectives are best solved with which kind of AI (or non-AI) approaches. Given the wealth of existing and newly released AI approaches, this is far from a trivial task.</t>

<section anchor="sub-challenge-suitable-approach-for-given-input"><name>Sub-challenge: Suitable Approach for Given Input</name>

<t>Different problems in network management come with widely different problem parameters. For example, security-related problems may have large amounts of textual or encrypted data as input, whereas forecasting problems have historical time series data as input. They also vary in the amount of available data.</t>

<t>Both the type and amount of data influences the selection of an appropriate AI technique. On one hand, in scenarios with small dimensional data, classical machine learning techniques (e.g., SVM, tree-based approaches, etc.) are often sufficient and even superior to NNs <xref target="Gre19"/>. On the other hand, NNs have the advantage of learning complex models from large amounts of data without requiring feature engineering. Here, different neural network architectures are useful for different kinds of problems. The traditional and simplest architecture are (fully connected) Multi-Layer Perceptrons (MLPs), which are useful for structured, tabular data. For images, videos, or other high-dimensional data with correlation between “close” features, convolutional neural networks (CNNs) are useful. Recurrent neural networks (RNNs), especially the Long Short-Term Memory (LSTM) architecture, and attention-based neural networks (transformers) are great for sequential data like time series or text. This evolution leads to the era of LLMs also impacting research in networking <xref target="Hua25"/><xref target="Wu24"/>.</t>

<t>It is worth noting that Graph Neural Networks (GNNs) can incorporate and consider the graph-structured input, which is very useful in network management <xref target="Jie22"/>, e.g., to represent the network topology.</t>

<t>The aforementioned rough guidelines can help identify a suitable AI approach or NN architecture. Still, best results are often achieved with a sophisticated combination of different approaches. For example, multiple elements can be combined into one architecture <xref target="Ham23"/>, e.g., with both CNNs and LSTMs, and multiple separate AI approaches can be used as an ensemble to combine their strengths <xref target="Das23"/>. Here, simplifying the mapping from problem type and input to suitable AI approaches and architectures is clearly an open challenge. Future work should address this challenge by providing both clearer guidelines and striving for more general AI approaches that can easily be applied to a large variety of different problem inputs.</t>

</section>
<section anchor="sub-challenge-suitable-approach-for-desired-output"><name>Sub-challenge: Suitable Approach for Desired Output</name>

<t>Similar to the challenge of identifying suitable AI approaches for a given problem input, the desired output for a given problem also affects which AI approach should be chosen. Here, the format of the desired output (single value, class, action, etc.), the frequency of these outputs and their meaning should be considered.</t>

<t>Again, there are rough guidelines for identifying a group of suitable AI approaches. For example, if a single numerical value is required (e.g., the amount of resources to allocate to a service instance), then a supervised regression approaches should be considered as a first candidate option as it is lightweight for this type of task. In case of classification (e.g., of malware or another security issue <xref target="Abd10"/>) instead of predicting a value is desired, supervised methods can be used if labelled training data is available. There are also cases where a single class of training data is available, as for example in the context of anomaly detection where the model is fitted to normal data. In that case, one-class supervised techniques can be considered as a good candidate. Alternatively, unsupervised machine learning can help to cluster given data into separate groups, which can be useful to analyse networking data, e.g., for better understanding different types of traffic or user segments. Furthermore, the quality of the data <xref target="Liu22"/> directly impacts on the robustness of a ML model with the risk of biased models due to over-fitting. As highlighted with these few examples, finding a suitable approach to a problem depends on many factors including the type of problem to handle but also other contextual elements such as the availability and the quality of data. To help in building AI-based solutions, pipeline generators have merged with automated capabilities, paving the way to the field of AutoML <xref target="Urb23"/>.</t>

<t>In addition to these classical supervised and unsupervised methods, Reinforcement Learning (RL) approaches allows active, sequential decisions rather than simple predictions or classification. This is often useful in network management, e.g., to actively control service scaling and placement <xref target="Sah23"/>. RL agents autonomously select suitable actions in a given environment and are especially useful for self-learning network management. In addition to model-free RL, model-based planning approaches (e.g., Monte Carlo Tree Search) also allow choosing suitable actions but require full knowledge of the environment dynamics. In contrast, model-free RL is ideal for scenarios with unknown environment dynamics, which is often the case in network management.</t>

<t>Like the previous sub-challenge, these are just rough guidelines that can help to select a suitable group of AI approaches. Identifying the most suitable approach within the group, e.g., the best out of the many existing reinforcement learning approaches, is still challenging. And, as before, different approaches could be combined to enable even more effective network management (e.g., heuristics + RL, LSTMs + RL, etc.). Here, further research can simplify the mapping from desired problem output to choosing or designing a suitable AI approach.</t>

</section>
<section anchor="tailoringAI"><name>Sub-challenge: Tailoring the AI Approach to the Given Problem</name>

<t>After addressing the two aforementioned sub-challenges, one may have selected a useful kind of AI approach for the given input and output of a network management problem. For example, one may select regression and supervised learning to forecast upcoming network traffic or select reinforcement learning to continuously control network and service coordination (scaling, placement, etc.). However, even within each of these fields (regression, reinforcement learning, etc.), there are many possible algorithms and hyper-parameters to consider. Selecting a suitable algorithm and parametrizing it with the right hyper-parameters is crucial to tailor the AI approach to the given network management problem.</t>

<t>For example, there are many different regression techniques (classical linear, polynomial regression, lasso/ridge regression, support vector regression, regression trees, neural networks, etc.), each with different benefits and drawbacks and each with its own set of hyper-parameters. Choosing a suitable technique depends on the amount and structure of the input data as well as on the desired output. It also depends on the available amount of compute resources and compute time until a prediction is required. If resources and time are not a limiting factor, many hyper-parameters can be tuned automatically.</t>

<t>This sub-challenge holds for all fields of AI: supervised learning (regression and classification), self-supervised learning, unsupervised learning, and reinforcement learning, each are broad and rapidly growing fields. Selecting suitable algorithms and hyper-parameters to tailor AI approaches to the network management problem is both an opportunity and a challenge. Here, future work should further explore these trade-offs and provide clearer guidelines on how to navigate these trade-offs for different network management tasks. As already mentioned, the AutoML field of research provides solutions to better customize ML algorithms and pipeline. However, such kind of optimization should be optimized according to domain-specific metrics rather than pure-AI metrics only. For instance, the integration of network-specific knowledge can be done through human feedback <xref target="Arz21"/>.</t>

</section>
</section>
<section anchor="performanceModels"><name>Performance of produced models</name>

<t>From a general point of view, any AI technique will produce results with a certain level of quality. This leads to four inherent questions: (1) what is the definition of the performance in a context of a NM application? (2) How to measure it? (3) How to ensure the quality of produced results by AI is aligned with NM objectives? (4) How to maintain or improve the quality of produced results?</t>

<t>Many metrics have been already defined to evaluate the performance of an AI-based technique according to its NM-level objectives. For example, QoS metrics (throughput, latency) can serve to measure the performance of a routing algorithm along with the computational complexity (memory consumption, size of routing tables). The question is to model and measure these two antagonist types of metrics. Number of true/false positives/negatives are the most basic metrics for network attack detection functions. Although the first two questions are thus already answered even if improvement can be done, question (3) refers to the integration of metrics into AI algorithms. Its objective is to obtain the best results which need to be quantified with these metrics. Depending on the type of algorithm, these metrics are either evaluated in an online manner with a feedback loop (for example with reinforcement learning) or in batch to optimize a model based on a particular context (for example described by a dataset for machine learning).</t>

<t>The problem is twofold. First, the performance can be measured through multiple metrics of different types (numerical or ordinal, for example), and some can be constrained by fixed boundaries (such as maximum latency), making their joint use challenging when creating an AI model to resolve an NM problem. Second, the scale of the metrics differs in importance or impact and can eventually vary across domains. It can be hard to precisely assess what a good or bad value is (as it might depend on multiple other ones) and it is even more difficult to integrate this in an AI technique, especially for learning algorithms to adjust their models based on performance. Indeed, many learning algorithms run through multiple iterations and rely on internal metrics (Mean Absolute Error (MAE) or Mean Squared Error (MSE) for neural network, Gini index or entropy for decision trees, distance to a hyperplane for SVMs, etc.) which are not strongly correlated to the final operational metrics of the NM application. AI-internal metrics such as the loss do not match well the metrics related to the final NM objectives, thus the significance and impact of the AI errors cannot be easily translated into the NM domain.</t>

<t>For instance, a decision tree algorithm for classification purposes aims to create branches with a maximum of data from the same classes and to avoid mixing classes. It is done thanks to a criterion like the entropy index, but this kind of index does not assume any difference between mixing classes A and B or A and C. Assuming an operational point of view in which A and B mixing in the predictions is not critical, the algorithm should have preferred to mix A and B rather than A and C, even if in the first case it will produce more errors. Therefore, the internal functioning of the AI algorithms should be refined, specifically by defining a particular criterion to replace entropy as a quality measure when separating two branches. It assumes that the final NM objectives are integrated at this stage.</t>

<t>Another concrete example is traffic predictors, which aim to forecast traffic demands. They produce an output that is not necessarily simple to interpret and use by, for example, capacity allocation strategies/policies. A traditional traffic prediction that tries to minimize (perfectly symmetric) MAE/MSE treats positive and negative errors identically, hence is agnostic to the diverse meaning (and costs) of under- and over-provisioning. And, such a prediction does not provide any information on, for example, how to dimension resources/capacity to accommodate future demand while avoiding all under-provisioning (which causes service disruption) while minimizing over-provisioning (i.e., wasting resources). In other words, it forces the operator to guess the over-provisioning by taking uninformed safety margins. A more sensible approach is to forecast directly the needed capacity, rather than the traffic <xref target="Beg19"/>.</t>

<t>While this is just an example, the high-level challenge is devising forecasting models that minimize the correct objective/loss function for the specific NM task at hand (instead of generic MAE/MSE). In this way, the prediction phase becomes an integral part of NM, and not just a limited and difficult-to-use input to it. In ML terms, this maps to solving the loss-metric mismatch in the context of anticipatory NM <xref target="Hua19"/>.</t>

<t>Another issue for statistical learning (from examples/observations) is mainly about extracting an estimator from a finite set of input-output samples drawn from an unknown probability distribution that should be descriptive enough for unseen/new input data. In this context, online monitoring and error control of the quality/properties of these point estimators (bias, variance, mean squared error, etc.) is critical for dynamic/uncertain network environments. Similar reasoning/challenge applies to interval estimates, i.e., confidence intervals (frequentist) and credible intervals (Bayesian).</t>

<t>Finally, question (4) refers to the ability of an AI solution to remain efficient and eventually improve over time <xref target="Ans26"/>. This requires dynamic methods capable to adapt to a changing environment. As already highlighted, the models can be dynamically adjusted based on the errors they produced. In the context of ML, the models can be also updated based on new data, either through a complete re-learning phase, fine-tuning or transfer-learning. This assumes to collect and ingest continuously new data. However, as highlighted in <xref target="Sha22"/>, this type of ML, qualified as online or incremental, raises several challenges when applied to traffic analysis. For example, there is a set of related challenges related to select or discard some data over a time horizon and to label data in real-time. Other challenges are more generic to this ML research area such as class imbalance or concept drift.</t>

</section>
<section anchor="lightweightAI"><name>Lightweight AI</name>

<t>Network management and operations often need to be performed under strict time constraints, i.e., at line rate, particularly in autonomic or self-driven networks. Locating NM functions as close as possible to where forwarding is achieved is an interesting option to avoid additional delays when these operations are performed remotely, for example in a centralized controller. Besides, forwarding devices may offer available resources to supplement or replace edge resources. For AI coupled with network management, AI tasks can be offloaded in network devices or more generally embedded within the network. Obviously, time-critical tasks are the best candidates to be offloaded within the network. Costly learning tasks should be processed in high-end servers but created models can be deployed, configured, modified and tuned in switches.</t>

<t>Recent advances in network programmability facilitate the programming of specific tasks at the data-plane level. P4 <xref target="Bos14"/> is widely used today for many tasks including firewalling <xref target="Dat18"/> or bandwidth management <xref target="Che19"/>. P4 is generally agnostic to specific hardware. It is based on the RMT (Reconfigurable Match Table) architectural model <xref target="Bos13"/> that is generally accepted to be generic enough to represent limited but essential switch architecture components and functionalities. The RMT model allows reconfiguring match-action tables where actions can be usual ones (rewrite some headers, forward, drop...). Actions are thus applied on the packets when they are forwarded. Actions can also be more complex programs with some safeguards: no loops, resistivity, etc. The impact on the program development is huge. For example, real number operations are not available by default while they are widely used in many AI algorithms.</t>

<t>In a nutshell, the first challenge to overcome in embedding AI in a network is the hardware's capacity to support AI operations (architectural limitation). Considering software equipment such as a virtual switch simplifies the problem but does not entirely resolve it, as even in that case, strict line-rate requirements limit the type of programs that can be executed. For example, BPF (Berkeley Packet Filter) <xref target="Mcc93"/> programs provide greater control over packet processing in OVS <xref target="Cha18"/> but still have some limitations, as the execution time of these programs is bounded by nature to ensure their termination, an essential requirement assuming the run-to-completion model which permits high throughput.</t>

<t>The second challenge of network-embedded AI in the network is allocating enough resources for AI tasks while minimizing impact on other network device tasks such as forwarding, monitoring, and filtering. Approximation and/or optimization of AI tasks are potential directions to help in this area. For instance, many network monitoring proposals rely on sketches with well-tuned implementations for the data-plane <xref target="Liu16"/><xref target="Yan18"/>. However, no general optimized AI-programmable abstraction exists to fit all cases, and proposals are mostly use-case centric. There have been many proposed approaches to develop specific P4 programs for many NM tasks, including involving ML. For each, this requires a specific adaptation <xref target="Hau23"/> with a few attempts to propose generic programs to be reusable or composed as a kind of libraries such as <xref target="Zha24"/> leveraging quantization, <xref target="Jos21"/><xref target="Jos22"/> relying on pre-computed lookup tables of real-value functions or <xref target="Swa23"/> proposing function templates for common operations. Besides, distributed processing is a common technique to distribute the load of a single task between multiple entities. AI task decomposition between network elements, edge servers, or controllers has also been proposed <xref target="Gup18"/>.</t>

</section>
<section anchor="DistributedAI"><name>Distributed AI</name>

<t>Distributed AI assumes different related tasks and components to be distributed across computational and possibly heterogeneous resources. For example, with advances in transfer and Federated Learning (FL), models can be learned, partially shared and combined or data can be also shared to either improve a local or global model. By nature, a network and a networked system is distributed and is thus well adapted to any distributed application. This is exacerbated with the deployment of fog infrastructure mixing network and computational resources. Hence, network management can directly benefit to the distributed network structure to solve its own particular problems, but any other type of AI-based distributed applications also assumes communication technologies to enable interactions between the different entities. This leads to the two sub-challenges described hereafter.</t>

<section anchor="netManFordistributedAI"><name>Network management for efficient distributed AI</name>

<t>Distributed AI relies on exchanging information between different entities and comes with various requirements in terms of volume, frequency, security, etc. This can be mapped to network requirements such as latency, bandwidth or confidentiality. Therefore, the network needs to provide adequate resources to support the proper execution of the AI distributed application. While this is true for any distributed application, the nature of the problem intended to be solved by an AI application and how this would be solved should be considered <xref target="Lin21"/>. For example, in the context of optimizing FL <xref target="Che22"/>, local models can be shared to create a global model. In case of failure of network links or too high latency, some local models might not be appropriately integrated into the global model with possible impact on AI performance. Depending on the nature of the latter, it might be better to guarantee high-performance communications with a few nodes or to ensure connectivity between all of them even with lower network performance. Mapping is thus necessary between the network management plane and the distributed AI applications which leads to a set of questions to be addressed about interfaces, data and information models or protocols. While the network can be adapted or eventually adapt itself to the AI distributed applications, AI applications could also adapt themselves to the underlying network conditions <xref target="Raj24"/>. It paves the way to research on methods to support AI-aware NM or network-aware AI applications or a combination of both.</t>

</section>
<section anchor="distributed-ai-for-network-management"><name>Distributed AI for network management</name>

<t>For network management applications relying on distributed AI, challenges from <xref target="netManFordistributedAI"></xref> are still valid. Furthermore, network management problems also consider network-specific elements like traffic to be analyzed or configuration to be set on distributed network equipment. Co-locating AI processing and these elements (fully or partially) may help increase performance. For example, pre-calculation on traffic data can be offloaded on network routers before being further processed in high-end servers in a data-center. In general, distributed AI-based network management decisions could be made at different nodes in the network based on locally available information <xref target="Sch21"/>. Hence, deployment of AI-based solutions for network management can also consider various network attributes like network topology, routing policies, or network device capability. In that case, management of computational and network resources is even more coupled than in <xref target="netManFordistributedAI"></xref>, since the network is both part of the AI pipeline resources and the managed object through AI.</t>

<t>A primary application for distributed AI is for management problems that have a local scope. One example concerns problems that can be addressed at the edge, involving tasks and control loops that monitor and apply local optimizations to the edge in isolation from activities conducted by other instances across the network.   However, distributed AI can involve techniques in which multiple entities collaborate to solve a global problem. Such solutions lend themselves to problems in which centralized solutions face certain foundational challenges such as security, privacy, and trust: The need to maintain complete state in a centralized solution may not be practical in some cases due to concerns such as privacy and trust among multiple subdomains, which may not want to share all their data even if they would be willing to collaborate on a problem. Other foundational challenges concern issues related to timeliness, in which distributed solutions may have inherent advantages over centralized solutions as they avoid delays caused by the need to communicate updates globally and across long distances.</t>

</section>
</section>
<section anchor="planningActions"><name>AI for planning of actions</name>

<t>Many tasks in network management revolve around the planning of actions with the purpose of optimizing a network and facilitating the delivery of communication services.  For example, communication paths need to be planned and set up in ways that minimize wasted network resources (to optimize cost) while facilitating high network utilization (avoiding bottlenecks and the formation of congestion hotspots) and ensuring resiliency (by making sure that backup paths are not congruent with primary paths).  Other examples were mentioned in <xref target="pbNM"></xref>.</t>

<t>The promise of central control is that decisions can be optimized when made with complete knowledge of relevant context, as opposed to distributed control, which needs to rely on local decisions made with incomplete knowledge while incurring higher overhead to replicate relevant state across multiple systems.  However, as the scale of networks and interconnected systems continues to grow, so does the size of the planning task. Many problems are NP-hard. As a result, solutions typically need to rely on heuristics and algorithms that often result in suboptimal outcomes and that are challenging to deploy in a scalable manner <xref target="Ahm21"/>.</t>

<t>The emergence of Intent-Based Networking (IBN) emphasizes the need for automated planning even further. IBN should allow users (network operators, not end users of communication services) to articulate desired outcomes without the need to specify how to achieve those outcomes. An Intent-Based System is responsible for translating the intent into courses of action that achieve the desired outcomes and that continue to maintain the outcomes over time <xref target="RFC9315"/>. How the necessary courses of action are derived and what planning needs to take place is left open, but that is where the real challenge lies.  Solutions that rely on clever algorithms devised by human developers face the same challenges as any other network management tasks.</t>

<t>These properties (problems with a clearly defined need, whose solution is faced with exploding search spaces and that today rely on algorithms and heuristics that in many cases result only suboptimal outcomes and significant limitations in scale) make automated planning of actions an ideal candidate for the application of AI-based solutions <xref target="Abd24"/>.</t>

<t>A much-publicized leap in AI has been the development of AlphaGo <xref target="Sil16"/>.  Instead of using AI to merely solve classification problems, AlphaGo has been successful in automatically deriving winning strategies for board games, specifically the game of Go which features a prohibitively large search space that was long thought to put the ability to play Go at a world class level beyond the reach of problems that AI could solve.   Among the remarkable aspects of Alpha Go is that it is able to identify winning strategies completely on its own, without needing those strategies to be taught or learned through observations assuming the system is aware of rules.</t>

<t>The challenge for AI in network management is hence, where is the equivalent of an Alpha Go that can be applied to network management (and networking) problems?  Specifically, better solutions are needed for solutions that automatically derive plans and courses of actions for network optimization and similar NP-hard problems, such as provided today with only limited effectiveness by controllers and management applications.</t>

<t>Although AI-based solutions for the automated planning of actions, including the automated identification of courses of action, have not been explored as much as classification <xref target="Sal20"/> problems, the quest for autonomous networks in the last decade and the advent of 5G and B5G have led to a rapid increase in proposed solutions, such as within the context of Zero Touch Management <xref target="Cor22"/>.</t>

<t>Also, the evaluation of AI algorithms to derive courses of actions is complex. Contrary to game playing, solutions need to be applied in the real world, where actions have real effects and consequences. Different orientations can be envisioned. First, incremental application of AI decisions with small steps can allow careful observation and detection of unexpected effects. This can be complemented with rollback techniques. Second, verification techniques can be leveraged to verify that decisions made by AI are maintained within safety bounds <xref target="Xin24"/>. Third, sandbox environments can be used but they should be representative of the real world. After progress in simulation and emulation, recent research advances lead to the definition of digital twins, which implies tight coupling between a real system and its digital twin to ensure parallel but synchronized execution <xref target="Wu21"/>. Alternatively, transfer learning techniques are another promising area to capitalize on ML models applicable to real world systems, for instance, to learn an intrusion detector that can be instantiated in multiple environments <xref target="Sha24"/><xref target="Ans23"/>. Generally, it is also an open problem to make the use of AI more acceptable, as highlighted in the dedicated section.</t>

</section>
</section>
<section anchor="netData"><name>Network data as input for ML algorithms</name>

<t>Many applications of AI take data as input. The quality of the outputs of ML-based techniques is highly dependent on the quality and quantity of data used for learning and also during inference. For example, as modern network infrastructures move toward higher speed and scale, they aim to support increasingly more demanding services with strict performance guarantees. These often require resource reconfigurations at runtime in response to emerging network events so that they can ensure reliable delivery at the expected performance level. Timely observation and detection of events is also of paramount importance for security purposes and can allow faster execution of remedy actions, thus leading to reduced service downtime.</t>

<t>Thus, the challenge of data management is multifaceted, as detailed in the next subsections.</t>

<section anchor="dataForAI"><name>Data for AI-based NM solutions</name>

<t>Assuming a network management application, the first problem to address is to define the data to be collected, which will be appropriate to obtain accurate results. This data selection can require defining problem-specific data or features (feature engineering). Furthermore, machine learning algorithms only work as desired when data to be analyzed respects given properties. Many methods rely on vector-based distances, which assume that the data encoded into the vector respects the underlying distance semantic. Taking the first n bytes of a packet as vectors and computing distances accordingly is possible, but does not embed the semantic of the information carried in the headers. A solution to circumvent the problem of network traffic representation is to transform traffic data into a specific format that can be easily handled by NN architectures, for instance by creating images analyzed by convolutional neural networks <xref target="Sha21"/>. Data can also be easily represented as graphs, because topologies or communication graphs would require less adaptation to be given as inputs to GNNs <xref target="Jie22"/>. As an example, graph-based representations are considered practically efficient due to their intrinsic structural similarities with network data, as illustrated in <xref target="Bar23"/> for attack detection.</t>

<t>Since data to handle can also be in a schema-free or eventually text-based format, one example could be the automated annotation of management intents provided in unstructured textual format (policies descriptions, specifications) to extract management entities and operations. For that purpose, suitable annotation models need to be built using existing NER (Named Entity Recognition) techniques usually applied for NLP. However, this shall be carefully crafted or specialized for network management (intent) language which indirectly relates back to the challenges of AI techniques for NM specified earlier. Today, with the progress on LLMs, different proposed approaches have emerged as for example in <xref target="Mek24"/><xref target="Fua24"/><xref target="Dze23"/>.<br />
Secondly, similar to the problem of mapping AI algorithms with NM problems in section <xref target="problemMapping"></xref>, data to be collected also depends on the NM problem to be solved. The mapping between the data sources and the problem is not straightforward as all dependencies or correlations are not known in advance and some might be be discovered by the AI algorithms themselves <xref target="For23"/>. In addition, the types of data to collect can vary over time to maintain the performance of an AI-based application or to adapt it to a new context when learned models are updated dynamically.</t>

<t>Thirdly, the behaviour of any network is not just derived from the events that can be directly observed, such as network traffic overload, but also from events occurring outside the environment of the network. The information provided by the detectors of such kinds of events, e.g. a natural incident (earthquake, storm), can be used to determine the adaptation of the network to avoid potential problems derived from such events. Those can be provided by big data sources as well as sensors of many kinds. The challenge related to this task is to process large amounts of data and associate it with the effects that those events have on the network. It is hard to determine the static and dynamic relation between the data provided by external sources and the specific implications it may have in networks. For instance, the effect of a “flash crowd” detected in an external source may require adapting the network service configuration to support is related workload. The objective is to complement a control-loop, as shown in <xref target="Mar18"/>, by including the specific AI engines into the decision components as well as the processes that close the loop, so the AI engine can receive feedback from the network in order to improve its own behaviour.</t>

</section>
<section anchor="dataCollection"><name>Data collection</name>

<t>Once defined, the second problem to address is the collection of data. Monitoring frameworks have been developed for many years such as IPFIX <xref target="RFC7011"/> and more recently with SDN-based monitoring solutions <xref target="Yu14"/><xref target="Ngu20"/>. However, going towards more AI for actions in network management supposes also to retrieve more than traffic related information. Actually, configuration information such as topologies, routing tables or security policies have been proven to be relevant in specific scenarios. As a result, many different technologies can be used to retrieve meaningful data. To support improved QoE, monitoring of the application layer is helpful but far from being easy with the heterogeneity of end-user applications and the wide use of encrypted channels. Monitoring techniques need to be reinvented through the definition of new techniques to extract knowledge from raw measurement <xref target="Bri19"/> or by involving end-users with crowd-sourcing <xref target="Hir15"/> and distributed monitoring. Also, the data-mesh concept <xref target="Mac22"/> proposes to classify data into three categories: source-aligned, aggregate and consumer-aligned. Source-aligned data are those related to the same operational domain, and it is important to correlate or aggregate them with higher planes: management-, control- and forwarding plane. An issue is the difference, not only in the nature of data, but in their volumes and their variety. Some may change rapidly over time (for example network traffic) while other may be quite stable (device state).</t>

<t>The collecting process requirements depend on the kind of processing. We can distinguish two major classes: batch/offline vs real-time/online processing. In particular, real-time monitoring tools are key in enabling dynamic resource management functions to operate on short reconfiguration cycles. However, maintaining an accurate view of the network state requires a vast amount of information to be collected and processed. While efficient mechanisms that extract raw measurement data at line rate have been recently developed, the processing of collected data is still a costly operation. This involves potentially sampling, evaluating and aggregating a vast amount of state information as a response to a diverse set of monitoring queries, before generating accurate reports. One difficult problem resides also in the availability of data as real-time data from different sources to be aggregated may not arrive at the same time requiring so some buffering techniques. Machine learning methods, e.g. based on regression, can be used to intelligently filter the raw measurements and thus reduce the volume of data to process. For example, in <xref target="Tan20"/> the authors proposed an approach in which the classifiers derived for this purpose (according to measurements on traffic properties) can achieve a threefold improvement in the query processing capability. A residual question is the storage of raw measurements. In fact, predicting the lifetime of data is challenging because their analysis may not be planned and triggered by a particular event (for example, an anomaly or attack). As a result, the provisioning of storage capacity can be hard.</t>

<t>In parallel to the continuously increasing dynamic behavior of networks and complexity of traffic, there is a trend towards more user traffic processing customization <xref target="RFC8986"/><xref target="Li19"/>. As a result, fine grained information about network element states is expected and new proposed approaches have emerged to collect on-path data or in-band network telemetry information <xref target="Tan21b"/>. These new approaches have been designed by introducing much flexibility and customization and could be helpful to be used in conjunction with AI applications. However, the seamless coupling of telemetry processes with packet forwarding requires careful definition of solutions to limit the overhead and the impact of the throughput while providing the necessary level of details. This shares commonalities with the lightweight AI challenge.</t>

</section>
<section anchor="usableData"><name>Usable data</name>

<t>Although all agree on the necessity to have more shared datasets, it is quite uncommon in practice. Data contains private or sensitive information and may not be shared because of the criticality of data (which can be used by ill-intentioned adversaries) or due to laws or regulations, even within the same company. To solve this issue, anonymization techniques <xref target="Dij18"/> can be enhanced to optimize the trade-off between valuable data and sensitive information (potential) leakage or reconstruction. 
Whatever the final user of data, regulations and laws impose rules on data management with potentially costly impact if they are not respected voluntarily or not. Defining a new monitoring framework should always consider security and privacy aspects, for example to let any user/customer or access/remove its own data with General Data Protection Regulation (GDPR) in EU. The challenge resides here in the capacity of qualifying what is critical or private information and the capacity for an adversary to reconstruct it from other sources of data. Hence AI/ML based solutions will require more data but also more administrative, legal and ethical procedures. Those can last long and so slow down the deployment of a new solution. In addition, this requires interaction with experts from different domains (e.g. AI engineer and a lawyer) for example to ensure by-design privacy in traffic analysis techniques <xref target="Wan22"/>. The integration of these non-technical constraints should be considered when defining new data to be collected or a new technique to collect data. However, knowing the final use of data is most of the time necessary for ethical and legal assessment which assumes that those considerations should be integrated from the early design of new AI-based solutions.</t>

<t>For supervised or semi-supervised training, having a labelled dataset is a prerequisite. It constitutes a major challenge as well. While probes exist to collect real network data, those data are typically unlabelled. This limits application of ML to unsupervised learning tasks (learning from data). Because manual labelling is a tedious task, one option is to leverage AI to guide humans. This may also support a better generalization of a learned model. Indeed, an underlying challenge is the genericity or coverage of the datasets. Labels encode values of an objective function, the challenge posed by the design of such tools is tremendous since for involving a M:N relationship: 1 data type may be associated to M objective function values and N data types may be associated to 1 objective function. As a result, most datasets used for research encode a single label for a particular application like attack label for datasets to be used in the context of intrusion detection or application type for network traffic used for classification where the value of a single dataset could be capitalized in several applications. More generally, in the context of intrusion detection, using raw data rather than pre-processed data as it is common in open dataset has been demonstrated to be inefficient <xref target="Dub24"/>.</t>

<t>Again, researchers need empirical (or at least realistic) datasets to validate their solutions. Unfortunately, as highlighted above, having such data from real deployments for various reasons (business secrets, privacy concerns, concerns that vulnerabilities are revealed by accident, raw unlabelled data, etc.) is tough. Even if such a dataset is available it might not be enough to convincingly validate a new algorithm. 
Instead of falling back to artificial testbed experiments or simulation, it would be useful to have the capability to generate datasets with characteristics that are not 100% identical but like the characteristics of one or more real datasets. Such synthetic networks can be used to validate new management algorithms, intrusion detection systems, etc. 
The usage of AI, Generative Adversarial Networks (GANs), in this area <xref target="Hui22"/> is not yet widespread and there are still many concerns that deter researchers, e.g. the fear of leaking sensitive information from the original dataset into the synthetic dataset. Furthermore, a major underlying challenge is to generate data realistic enough. This requires formalizing and integrating network-specific constraints, such as protocol specification, when generating data <xref target="Jia24"/> or using a real testbed guided by AI to create challenging scenarios for validation purposes <xref target="Ans25"/>.</t>

</section>
</section>
<section anchor="acceptability"><name>Acceptability of AI</name>

<t>Networks are critical infrastructures. On one hand, they should be operated without interruption and must be interoperable. Networks, except in a lab, are not isolated, which slow down innovation in general. For example, changing Internet routing protocols need to be accepted by multiple entities such as operators of interconnected networks. The same applies for protocols. Even if there have been several versions of major protocols in use like TCP or DNS, there are still some security issues which cannot be patched with 100% guarantee. On the other hand, results provided by AI solutions are uncertain by nature. The same technique applied in different environments can produce different results. AI techniques need some effort (time and human) to be properly configured or to be stabilized. For instance, RL needs several iterations before being able to produce acceptable results. These properties of AI techniques are thus a bit antagonist with the criticality of network infrastructures. With that in mind, acceptability of AI by network operators is clearly an obstacle for its larger adoption.</t>

<section anchor="explainability"><name>Explainability of Network-AI products</name>

<t>A common issue across many ML applications is their lack to provide human-understandable reasoning processes. This means that, after training, the knowledge acquired by ML models is unintelligible to humans. As a result, offering hard guarantees on performance is a very challenging issue. In addition, complex ML models like neural networks -that often have more than hundreds of thousands of parameters- are very hard to debug or troubleshoot in case of failure.</t>

<t>While this is a common issue for all applications of AI, many areas work well with uncertainty and the black-box behaviour of AI-based solutions. For instance, users accept an inherent error in recommender systems or computer vision solutions.</t>

<t>The networking field has already produced a set of well-established network management algorithms and methods, with clear performance guarantees and troubleshooting mechanisms <xref target="Rex06"/><xref target="Kr15"/>. As such, improving debugging, troubleshooting and guarantees on AI-based solutions for networking is a must. AI researchers and practitioners are devoting large research efforts to improve this aspect of ML models, which is commonly known as explainability <xref target="XAI"/>.</t>

<t>This set of techniques provides insights and, in some cases, guarantees on the performance and behaviour of ML-based solutions. Understanding such techniques, researching and applying them to network AI is critical for the success of the field.</t>

<t>There exist several ML-based methods that are human-understandable, although not widely used today. For instance, <xref target="Mar20"/> shows a method for building anticipation models (prediction) that provide explanations while determining some actions for tuning some parameters of the network. There are other challenges that should be addressed, such as providing explanations for other ML methods that are quite extended. For instance, xNN/SVM models can be accompanied by Digital Twins of the network that are reversely explored to explain some output from the ML model (e.g., xNN/SVM). In this context, there already exist several methods <xref target="Men20"/><xref target="Puj21"/> that produce human-readable interpretations of trained NN models, by analysing their neural activations on different inputs. It is worth noting that Digital Twins are not considered per se an AI approach; they merely serve to provide a digital representation of a network that can serve as its proxy and offer a layer of indirection between management applications and actual network resources. However, it is conceivable that AI-based management applications can be combined and operate in conjunction with Digital Twin technology, for example to use a Digital Twins as an experimentation sandbox or staging ground for AI-driven applications.</t>

</section>
<section anchor="productionSystems"><name>AI-based products and algorithms in production systems</name>

<t>AI-based network management and optimization algorithms are first trained, then the resulting model is used to produce relevant inferences in operation, either in management or optimization scenarios. A relevant question for the success of AI-based solutions is: where does this training occur?</t>

<t>Traditionally, AI-based models have been trained in the same scenario where they operate<xref target="Val17"/><xref target="Xu18"/>, this is the customer network. However, this presents critical drawbacks. First, training an AI model for management and operation typically requires generating network configurations and scenarios that can break the network. This is because training requires seeing a broad spectrum of scenarios. Thus, training in production networks is very challenging. Second, customer networks may not be equipped with the monitoring infrastructure required to collect the data used in the training process (e.g., performance metrics). Performing learning directly into a production network is possible assuming imperfect models and the need of several step of refinement before it gets stable. For non-critical management task, such assumption can hold, and additional safe-guarding mechanisms should be considered in order to keep outputs of ML algorithms (such as decisions) within acceptable boundaries.</t>

<t>A more sensible approach is to train the AI-based product in a lab, for instance in the vendor’s premises. In the lab, AI models can be trained in a controlled testbed, with any configuration, even ones that break the network. However, the main challenge here arises from the fundamental differences between the lab’s network and the customer networks. For instance, the topology of the lab’s network might be smaller, etc. As a result, there is a need for the learned models to generalize. In this context, generalization means that models should be able to operate in other scenarios not seen during training, with different topologies, routing configurations, scheduling policies, etc. This well-known problem is due to the dependency between training and testing data when benchmarking models.</t>

<t>In order to address this generalization problem, multiple complementary approaches are possible:  One approach is training on diverse data that represents large parts of the expected problem space. For example, training with various traffic patterns will help improving the generalization of an AI-based traffic analyser. 
Another approach is to leverage AI designs or architectures that facilitate generalization. One example is GNNs <xref target="gnn1"/><xref target="gnn2"/>. They are a type of neural networks able to operate and generalize over graphs. Indeed, networks are fundamentally represented as graphs: topology, routing, etc. With GNNs, vendors can train the AI model in a lab with a certain topology and then directly use the resulting model in different customer networks, even with different network topologies. 
Finally, another approach is Transfer Learning <xref target="tl1"/>. With this technique, the knowledge gained in the lab’s training is used to operate in the customer network. Transfer Learning still requires that some data from the customer is used to re-train and fine-tune the model (e.g., accurate performance measurements). This means that, for each customer network, re-training is required. This may be problematic since it requires added cost and access to customer data.</t>

<t>In addition to the challenge of generalizing from training to production environment, there are also challenges in terms of interoperability between different AI approaches and different deployment environments. 
As mentioned above, AI approaches may be deployed in diverse environments, e.g., for training and production, but also for local development, for testing, and for validation or in different part of the production systems. 
These environments may differ in available compute resources, network topology, operating systems, cloud providers, etc. (single node machine, single cluster, many distributed clusters, etc.). Deploying the same AI solutions in these different environments can lead to various challenges in terms of interoperability. Common AI frameworks support scaling across networks of different size. Yet, many frameworks are often combined, e.g., for data collection, processing, predictions, validation, etc. Again, ensuring interoperability between these frameworks can be tedious.</t>

<t>This shares similarities with problems described in <xref target="DistributedAI"></xref> and particularly emphasizes the need for network environments to provide interfaces and descriptions suitable for AI solutions to be properly instantiated and configured.</t>

<t>One approach to address these interoperability challenges is through meta-frameworks that interface with most available AI frameworks. These meta-frameworks provide a higher level of abstraction and often allow seamless deployment across different environments (e.g., on-premises or at different cloud providers) <xref target="Mor18"/>.</t>

</section>
<section anchor="humansInLoop"><name>AI with humans in the loop</name>

<t>Depending on the network management task, AI can automate and replace manual human control, or it can complement human experts and keep them in the loop. Keeping humans in the loop will be an important step of building trust in AI approaches and help ensure the desired outcomes. 
There are various ways of keeping humans in the loop in the different fields of AI <xref target="Wu22"/>, which could be useful for different aspects of network management.</t>

<t>In classification tasks (e.g., detecting security breaches, malware or detecting anomalies), trained AI models provide a confidence score in addition to the predicted class. If the confidence is high, the prediction is used directly. If the confidence is too low, a human expert may jump in and make the decision - thereby also providing valuable training data to improve the AI model. Such approaches are already being used in industry, e.g., to automatically label datasets (AWS SageMaker). Similar approaches could also be used for other supervised learning tasks, e.g., regression. Still, it is an open challenge to keep humans in the loop in all phases of the learning process.</t>

<t>When using RL, e.g. to control service scaling and placement or route traffic flows, the agents typically interact with the environment (i.e., the simulated or real network) completely autonomously without human feedback. However, there is a growing number of approaches to put human experts back into the loop. One approach is offline reinforcement learning, where the training data does not come from the reinforcement learning agent’s own exploration but from pre-recorded traces of human experts (e.g., placement decisions that were made by humans before). Another approach is to reward the RL agent based on human feedback rather than a pre-defined reward function <xref target="Lee21"/>. Again, while there are first promising approaches, more work is required in this area. 
Overall, it is an open challenge to both leverage the benefits of AI but keep human experts in the loop where it is useful.</t>

</section>
</section>
<section anchor="conclusions"><name>Conclusions and Future Research Directions</name>

<t>This document has established a six-characteristic taxonomy to map the inherent architectural challenges found at the intersection of Network Management and Artificial Intelligence. To prevent this research from remaining siloed in standalone academic configurations, the NMRG and the broader NM community must focus technical resources on resolving the friction points where abstract AI/ML models meet deterministic network operations.</t>

<t>Key takeaways include:</t>

<t><list style="symbols">
  <t>NM problems often require network-specific solutions rather than direct imports from other AI domains.</t>
  <t>Data quality, representation, and privacy-aware sharing are fundamental enablers for effective AI-based NM.</t>
  <t>Lightweight and distributed AI architectures are needed to meet timing, resource, and deployment constraints.</t>
  <t>Seamless integration with existing NM processes, explainability, and human-in-the-loop interaction are essential for acceptability.</t>
  <t>Practical deployment also requires evaluation methods, safeguards, and mechanisms for models to generalize and adapt over time.</t>
</list></t>

</section>
<section anchor="security-considerations"><name>Security considerations</name>

<t>This document introduces the challenges of coupling AI and NM. Since the aim of this document is not to address a particular NM problem by defining a solution and because many possible ones can be developed further, it is not possible at this stage to define security concerns specific to a solution. However, examples of applications mentioned and cited in the different sections may face their own security concerns. In this section, our objective is to highlight high-level security considerations to be considered when coupling AI and NM. Those concerns serve as the common basis to be refined according when a particular NM application is developed.</t>

<section anchor="ai-based-security-solutions"><name>AI-based security solutions</name>

<t>The first security consideration refers to the use of AI for NM problem related to security of the managed networked systems. There are multiple scenarios where AI can be leveraged: to perform traffic filtering, to detect anomalies or to decide on target moving defence strategies. In these cases, the performance of the AI algorithms impacts on the security performance (e.g. detection or mitigation effectiveness) like any other non-AI system. However, AI methods generally tend to obfuscate how predictions are made and decisions taken. Explainability of AI is thus highly important and is addressed globally in section <xref target="explainability"></xref>.</t>

<t>Assuming a ML trained model, there is always an uncertainty regarding reachable performance on the wild once the solution is deployed as it can suffer from a poor generalization due to different reasons. There are two major problems which are well known in the ML field: overfitting or under-fitting.  In the first case, the learned model is too specific to the training data while in the second case the model does not infer any valuable knowledge from data. To avoid these issues, hyper-parameter fine-tuning is necessary. For example, the number of iterations is an essential hyper-parameter to be adjusted to learn a neural network model. If it is too low, the learning does not converge to a representative model of the training data (underfitting). With a high value, there is a risk that the model is too close to the training data (overfitting). In general, finding the right hyper-parameters is helpful to find a good ML algorithm configuration. There are different techniques ranging from grid-search to Bayesian optimization falling into the AI area of Hyper-Parameter Optimization (HPO) <xref target="Bis23"/>. This consideration goes beyond the sole problem of hyper-parameters settings but as a full analysis pipeline also assumes the ML algorithm to be selected or the data pre-processing to be configured. This reflects challenges from the AI research area covered by AutoML technique. In this document, this is also referred in section <xref target="tailoringAI"></xref> when considered in the context of NM. As highlighted in the aforementioned section, some expertise or area-specific knowledge can help guiding automated configuration processes.</t>

<t>Besides, machine learning assumes to have representative training data. The quality of dataset for learning is a vast problem. Additionally, the representation of data needs to be addressed carefully to be properly analysed by AI models, for example with pre-processing techniques to normalize data, balance classes or encode categorical features depending on the type of algorithms. Actually, section <xref target="netData"></xref> of this document fully addresses the concerns related to data in regard to NM problems.</t>

</section>
<section anchor="securityOfAI"><name>Security of AI</name>

<t>Although ensuring a good performance of AI algorithms is already challenging, assuming an attacker aiming at compromising it emphasizes the problem. Adversarial AI and notably adversarial ML have attracted a lot of attention over the last year. Adversarial AI and ML relates to both attack and defences. While this is out of scope of the document, evaluating threats against an ML system before deploying it is an important aspect. This supposes to assess what types of information the attacker can access (training data, trained model, algorithm configuration...) and the performable malicious actions (inject false training data, test the system, poison a model, etc.) to evaluate the magnitude of the impact of possible attacks.</t>

<t>For illustration purposes, we refer hereafter to some examples. In the case of an intrusion detector, an attacker may try to poison training data by providing adversarial samples to ensure that the detector will miss-classify the future attacks <xref target="Jmi22"/>. In a white-box approach where the model is known from the attacker, the attack can be carefully crafted to avoid being properly labelled. For instance, packet sizes and timings can be easily modified to bypass ML-based traffic classification system <xref target="Nas21"/>. In a black-box model, the attacker ignores the functioning or training data of the ML systems but can try to infer some information. For example, the attacker can try to reconstruct sensitive information which have been used for training. This type of attack qualified as model inversion <xref target="Fre15"/> raises concern regarding privacy.</t>

<t>All these threats are exacerbated in the context of solutions relying on distributed AI involving multiple entities that are not necessarily controlled by the same authority. Once the threats are assessed, solutions need to be developed and deployed which can be either proactive by providing some guarantees regarding the involved entities using authentication or trust mechanisms but also reactive by validating data processing through voting mechanisms or knowledge proofs. Other solutions include defensive techniques to rate limit or filter queries to a particular deployed model. All these examples are for illustration purposes and are not exhaustive.</t>

</section>
<section anchor="relevanceOfAIOutput"><name>Relevance of AI-based outputs</name>

<t>Security breaches can be created by an AI-driven application. Generally, any system that will be used to guide or advice on actions to be perform on network raise the same issue. For example, if an AI algorithm decides to change the filtering tables in a network it may compromise access control policies. Irrelevant results could be also produced. In the area of QoS, an AI system could allocate a bandwidth to a flow higher to the real link capacity. As shown from these two examples, an AI can produce decisions or values which are out of bounds of normal operations. To avoid such issues, safeguards can be added to discard or correct irrelevant outputs. Detecting such type of outputs can be also challenging in complex and distributed systems such as a network. Formal verification methods or testing techniques are helpful in that context.</t>

</section>
</section>
<section anchor="iana-considerations"><name>IANA Considerations</name>

<t>This document has no IANA actions.</t>

</section>


  </middle>

  <back>


    <references title='Normative References'>

&RFC7011;
&RFC8986;
&RFC9315;


    </references>

    <references title='Informative References'>

<reference anchor="Mac22" >
  <front>
    <title>Data Mesh: Concepts and Principles of a Paradigm Shift in Data Architectures</title>
    <author initials="I. A." surname="Machado" fullname="Inês Araújo Machado">
      <organization></organization>
    </author>
    <author initials="C." surname="Costa" fullname="Carlos Costa">
      <organization></organization>
    </author>
    <author initials="M. Y." surname="Santos" fullname="Maribel Yasmina Santos">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.procs.2021.12.013"/>
<annotation>Procedia Computer Science, vol. 196, pp. 263-271</annotation></reference>
<reference anchor="Wu22" >
  <front>
    <title>A survey of human-in-the-loop for machine learning</title>
    <author initials="X." surname="Wu" fullname="Xingjiao Wu">
      <organization></organization>
    </author>
    <author initials="L." surname="Xiao" fullname="Luwei Xiao">
      <organization></organization>
    </author>
    <author initials="Y." surname="Sun" fullname="Yixuan Sun">
      <organization></organization>
    </author>
    <author initials="J." surname="Zhang" fullname="Junhang Zhang">
      <organization></organization>
    </author>
    <author initials="T." surname="Ma" fullname="Tianlong Ma">
      <organization></organization>
    </author>
    <author initials="L." surname="He" fullname="Liang He">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.future.2022.05.014"/>
<annotation>Elsevier Future Generation Computer Systems, vol. 135, pp. 364-381</annotation></reference>
<reference anchor="Dub24" >
  <front>
    <title>Faulty use of the CIC-IDS 2017 dataset in information security research</title>
    <author initials="R." surname="Dube" fullname="R. Dube">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s11416-023-00509-7"/>
<annotation>Journal of Computer Virology and Hacking Techniques, 20(1), pp. 203-211</annotation></reference>
<reference anchor="Jia24" >
  <front>
    <title>NetDiffusion: Network Data Augmentation Through Protocol-Constrained Traffic Generation</title>
    <author initials="X." surname="Jiang" fullname="Xi Jiang">
      <organization></organization>
    </author>
    <author initials="S." surname="Liu" fullname="Shinan Liu">
      <organization></organization>
    </author>
    <author initials="A." surname="Gember-Jacobson" fullname="Aaron Gember-Jacobson">
      <organization></organization>
    </author>
    <author initials="A. N." surname="Bhagoji" fullname="Arjun Nitin Bhagoji">
      <organization></organization>
    </author>
    <author initials="P." surname="Schmitt" fullname="Paul Schmitt">
      <organization></organization>
    </author>
    <author initials="F." surname="Bronzino" fullname="Francesco Bronzino">
      <organization></organization>
    </author>
    <author initials="N." surname="Feamster" fullname="Nick Feamster">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3639037"/>
<annotation>Proceedings of the ACM on Measurement and Analysis of Computing Systems (POMACS), 8(1)</annotation></reference>
<reference anchor="Wan22" >
  <front>
    <title>Multiple Strategies Differential Privacy on Sparse Tensor Factorization for Network Traffic Analysis in 5G</title>
    <author initials="J." surname="Wang" fullname="J. Wang">
      <organization></organization>
    </author>
    <author initials="H." surname="Han" fullname="H. Han">
      <organization></organization>
    </author>
    <author initials="H." surname="Li" fullname="H. Li">
      <organization></organization>
    </author>
    <author initials="S." surname="He" fullname="S. He">
      <organization></organization>
    </author>
    <author initials="P. K." surname="Sharma" fullname="P. Kumar Sharma">
      <organization></organization>
    </author>
    <author initials="L." surname="Chen" fullname="L. Chen">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TII.2021.3082576"/>
<annotation>IEEE Transactions on Industrial Informatics, 18(3), pp. 1939-1948</annotation></reference>
<reference anchor="For23" >
  <front>
    <title>ADT: AI-Driven network Telemetry processing on routers</title>
    <author initials="P." surname="Foroughi" fullname="Parisa Foroughi">
      <organization></organization>
    </author>
    <author initials="F." surname="Brockners" fullname="Frank Brockners">
      <organization></organization>
    </author>
    <author initials="J.-L." surname="Rougier" fullname="Jean-Louis Rougier">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comnet.2022.109474"/>
<annotation>Elsevier Computer Networks, vol. 220</annotation></reference>
<reference anchor="Bar23" >
  <front>
    <title>FTG-Net: Hierarchical Flow-to-Traffic Graph Neural Network for DDoS Attack Detection</title>
    <author initials="L." surname="Barsellotti" fullname="L. Barsellotti">
      <organization></organization>
    </author>
    <author initials="L. D." surname="Marinis" fullname="L. De Marinis">
      <organization></organization>
    </author>
    <author initials="F." surname="Cugini" fullname="F. Cugini">
      <organization></organization>
    </author>
    <author initials="F." surname="Paolucci" fullname="F. Paolucci">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/HPSR57248.2023.10147929"/>
<annotation>IEEE International Conference on High Performance Switching and Routing (HPSR)</annotation></reference>
<reference anchor="Dze23" >
  <front>
    <title>LLM-Based Policy Generation for Intent-Based Management of Applications</title>
    <author initials="K." surname="Dzeparoska" fullname="K. Dzeparoska">
      <organization></organization>
    </author>
    <author initials="J." surname="Lin" fullname="J. Lin">
      <organization></organization>
    </author>
    <author initials="A." surname="Tizghadam" fullname="A. Tizghadam">
      <organization></organization>
    </author>
    <author initials="A." surname="Leon-Garcia" fullname="A. Leon-Garcia">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.23919/CNSM59352.2023.10327837"/>
<annotation>International Conference on Network and Service Management (CNSM)</annotation></reference>
<reference anchor="Fua24" >
  <front>
    <title>An Intent-based Networks Framework based on Large Language Models</title>
    <author initials="A." surname="Fuad" fullname="A. Fuad">
      <organization></organization>
    </author>
    <author initials="A. H." surname="Ahmed" fullname="A. H. Ahmed">
      <organization></organization>
    </author>
    <author initials="M. A." surname="Riegler" fullname="M. A. Riegler">
      <organization></organization>
    </author>
    <author initials="T." surname="Čičić" fullname="T. Čičić">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/NetSoft60951.2024.10588879"/>
<annotation>IEEE International Conference on Network Softwarization (NetSoft)</annotation></reference>
<reference anchor="Mek24" >
  <front>
    <title>Intent-Based Management of Next-Generation Networks: an LLM-Centric Approach</title>
    <author initials="A." surname="Mekrache" fullname="A. Mekrache">
      <organization></organization>
    </author>
    <author initials="A." surname="Ksentini" fullname="A. Ksentini">
      <organization></organization>
    </author>
    <author initials="C." surname="Verikoukis" fullname="C. Verikoukis">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/MNET.2024.3420120"/>
<annotation>IEEE Network, 38(5), pp. 29-36</annotation></reference>
<reference anchor="Ans23" >
  <front>
    <title>Auto-tuning of Hyper-parameters for Detecting Network Intrusions via Meta-learning</title>
    <author initials="O." surname="Anser" fullname="Omar Anser">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="I." surname="Chrisment" fullname="Isabelle Chrisment">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/NOMS56928.2023.10154381"/>
<annotation>IEEE/IFIP Network Operations and Management Symposium (NOMS) - AnNet workshop</annotation></reference>
<reference anchor="Ans25" >
  <front>
    <title>TATA: Benchmark NIDS Test Sets Assessment and Targeted Augmentation</title>
    <author initials="O." surname="Anser" fullname="Omar Anser">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="I." surname="Chrisment" fullname="Isabelle Chrisment">
      <organization></organization>
    </author>
    <author initials="D." surname="Kondo" fullname="Daishi Kondo">
      <organization></organization>
    </author>
    <date year="2025"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/978-3-032-07884-1_2"/>
<annotation>European Symposium on Research in Computer Security (ESORICS), LNCS, pp. 21-41</annotation></reference>
<reference anchor="Ans26" >
  <front>
    <title>An Empirical Study on Configuration Robustness of Unsupervised IDS</title>
    <author initials="O." surname="Anser" fullname="Omar Anser">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="I." surname="Chrisment" fullname="Isabelle Chrisment">
      <organization></organization>
    </author>
    <date year="2026"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/NOMS69089.2026.11668115"/>
<annotation>IEEE/IFIP Network Operations and Management Symposium (NOMS)</annotation></reference>
<reference anchor="Sha24" >
  <front>
    <title>DTL-IDS: An optimized Intrusion Detection Framework using Deep Transfer Learning and Genetic Algorithm</title>
    <author initials="S." surname="Latif" fullname="Shahid Latif">
      <organization></organization>
    </author>
    <author initials="W." surname="Boulila" fullname="Wadii Boulila">
      <organization></organization>
    </author>
    <author initials="A." surname="Koubaa" fullname="Anis Koubaa">
      <organization></organization>
    </author>
    <author initials="Z." surname="Zou" fullname="Zhuo Zou">
      <organization></organization>
    </author>
    <author initials="J." surname="Ahmad" fullname="Jawad Ahmad">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.jnca.2023.103784"/>
<annotation>Elsevier Journal of Network and Computer Applications, vol. 221</annotation></reference>
<reference anchor="Wu21" >
  <front>
    <title>Digital Twin Networks: A Survey</title>
    <author initials="Y." surname="Wu" fullname="Y. Wu">
      <organization></organization>
    </author>
    <author initials="K." surname="Zhang" fullname="K. Zhang">
      <organization></organization>
    </author>
    <author initials="Y." surname="Zhang" fullname="Y. Zhang">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/JIOT.2021.3079510"/>
<annotation>IEEE Internet of Things Journal, 8(18), pp. 13789-13804</annotation></reference>
<reference anchor="Xin24" >
  <front>
    <title>Occam: A Programming System for Reliable Network Management</title>
    <author initials="J." surname="Xing" fullname="Jiarong Xing">
      <organization></organization>
    </author>
    <author initials="K.-F." surname="Hsu" fullname="Kuo-Feng Hsu">
      <organization></organization>
    </author>
    <author initials="Y." surname="Xia" fullname="Yiting Xia">
      <organization></organization>
    </author>
    <author initials="Y." surname="Cai" fullname="Yan Cai">
      <organization></organization>
    </author>
    <author initials="Y." surname="Li" fullname="Yanping Li">
      <organization></organization>
    </author>
    <author initials="Y." surname="Zhang" fullname="Ying Zhang">
      <organization></organization>
    </author>
    <author initials="A." surname="Chen" fullname="Ang Chen">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3627703.3650086"/>
<annotation>ACM European Conference on Computer Systems (EuroSys)</annotation></reference>
<reference anchor="Cor22" >
  <front>
    <title>Zero Touch Management: A Survey of Network Automation Solutions for 5G and 6G Networks</title>
    <author initials="E." surname="Coronado" fullname="Estefanía Coronado">
      <organization></organization>
    </author>
    <author initials="R." surname="Behravesh" fullname="Rasoul Behravesh">
      <organization></organization>
    </author>
    <author initials="T." surname="Subramanya" fullname="Tejas Subramanya">
      <organization></organization>
    </author>
    <author initials="A." surname="Fernàndez-Fernàndez" fullname="Adriana Fernàndez-Fernàndez">
      <organization></organization>
    </author>
    <author initials="M. S." surname="Siddiqui" fullname="Muhammad Shuaib Siddiqui">
      <organization></organization>
    </author>
    <author initials="X." surname="Costa-Pérez" fullname="Xavier Costa-Pérez">
      <organization></organization>
    </author>
    <author initials="R." surname="Riggio" fullname="Roberto Riggio">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/COMST.2022.3212586"/>
<annotation>IEEE Communications Surveys &amp; Tutorials, 24(4), pp. 2535-2578</annotation></reference>
<reference anchor="Sal20" >
  <front>
    <title>A review on machine learning–based approaches for Internet traffic classification</title>
    <author initials="O." surname="Salman" fullname="Ola Salman">
      <organization></organization>
    </author>
    <author initials="I. H." surname="Elhajj" fullname="Imad H. Elhajj">
      <organization></organization>
    </author>
    <author initials="A." surname="Kayssi" fullname="Ayman Kayssi">
      <organization></organization>
    </author>
    <author initials="A." surname="Chehab" fullname="Ali Chehab">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s12243-020-00770-7"/>
<annotation>Springer Annals of Telecommunications, 75(11-12), pp. 673-710</annotation></reference>
<reference anchor="Abd24" >
  <front>
    <title>A dynamic AI-based algorithm selection for Virtual Network Embedding</title>
    <author initials="A." surname="Bouroudi" fullname="Abdelmounaim Bouroudi">
      <organization></organization>
    </author>
    <author initials="A." surname="Outtagarts" fullname="Abdelkader Outtagarts">
      <organization></organization>
    </author>
    <author initials="Y." surname="Hadjadj-Aoul" fullname="Yassine Hadjadj-Aoul">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s12243-024-01040-6"/>
<annotation>Springer Annals of Telecommunications, 80(3-4), pp. 265-281</annotation></reference>
<reference anchor="Ahm21" >
  <front>
    <title>Scalability, Consistency, Reliability and Security in SDN Controllers: A Survey of Diverse SDN Controllers</title>
    <author initials="S." surname="Ahmad" fullname="S. Ahmad">
      <organization></organization>
    </author>
    <author initials="A. H." surname="Mir" fullname="A. H. Mir">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s10922-020-09575-4"/>
<annotation>Springer Journal of Network and Systems Management, 29(1)</annotation></reference>
<reference anchor="Che22" >
  <front>
    <title>Federated Learning Over Wireless IoT Networks With Optimized Communication and Resources</title>
    <author initials="H." surname="Chen" fullname="H. Chen">
      <organization></organization>
    </author>
    <author initials="S." surname="Huang" fullname="S. Huang">
      <organization></organization>
    </author>
    <author initials="D." surname="Zhang" fullname="D. Zhang">
      <organization></organization>
    </author>
    <author initials="M." surname="Xiao" fullname="M. Xiao">
      <organization></organization>
    </author>
    <author initials="M." surname="Skoglund" fullname="M. Skoglund">
      <organization></organization>
    </author>
    <author initials="H. V." surname="Poor" fullname="H. V. Poor">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/JIOT.2022.3151193"/>
<annotation>IEEE Internet of Things Journal, 9(17)</annotation></reference>
<reference anchor="Lin21" >
  <front>
    <title>Optimizing AI Service Placement and Resource Allocation in Mobile Edge Intelligence Systems</title>
    <author initials="Z." surname="Lin" fullname="Z. Lin">
      <organization></organization>
    </author>
    <author initials="S." surname="Bi" fullname="S. Bi">
      <organization></organization>
    </author>
    <author initials="Y. -J. A." surname="Zhang" fullname="Y. -J. A. Zhang">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TWC.2021.3081991"/>
<annotation>IEEE Transactions on Wireless Communications, 20(11)</annotation></reference>
<reference anchor="Raj24" >
  <front>
    <title>CASSINI: Network-Aware Job Scheduling in Machine Learning Clusters</title>
    <author initials="S." surname="Rajasekaran" fullname="Sudarsanan Rajasekaran">
      <organization></organization>
    </author>
    <author initials="M." surname="Ghobadi" fullname="Manya Ghobadi">
      <organization></organization>
    </author>
    <author initials="A." surname="Akella" fullname="Aditya Akella">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
<annotation>USENIX Symposium on Networked Systems Design and Implementation (NSDI)</annotation></reference>
<reference anchor="Swa23" >
  <front>
    <title>Homunculus: Auto-Generating Efficient Data-Plane ML Pipelines for Datacenter Networks</title>
    <author initials="T." surname="Swamy" fullname="T. Swamy">
      <organization></organization>
    </author>
    <author initials="A." surname="Zulfiqar" fullname="A. Zulfiqar">
      <organization></organization>
    </author>
    <author initials="L." surname="Nardi" fullname="L. Nardi">
      <organization></organization>
    </author>
    <author initials="M." surname="Shahbaz" fullname="M. Shahbaz">
      <organization></organization>
    </author>
    <author initials="K." surname="Olukotun" fullname="K. Olukotun">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3582016.3582022"/>
<annotation>ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS)</annotation></reference>
<reference anchor="Jos21" >
  <front>
    <title>InREC: In-network REal Number Computation</title>
    <author initials="M." surname="Jose" fullname="Matthews Jose">
      <organization></organization>
    </author>
    <author initials="K." surname="Lazri" fullname="Kahina Lazri">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="O." surname="Festor" fullname="Olivier Festor">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
<annotation>IFIP/IEEE International Symposium on Integrated Network Management (IM)</annotation></reference>
<reference anchor="Zha24" >
  <front>
    <title>A Machine Learning-Based Toolbox for P4 Programmable Data-Planes</title>
    <author initials="K." surname="Zhang" fullname="K. Zhang">
      <organization></organization>
    </author>
    <author initials="N." surname="Samaan" fullname="N. Samaan">
      <organization></organization>
    </author>
    <author initials="A." surname="Karmouch" fullname="A. Karmouch">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TNSM.2024.3402074"/>
<annotation>IEEE Transactions on Network and Service Management, 21(4), pp. 4450-4465</annotation></reference>
<reference anchor="Hau23" >
  <front>
    <title>A survey on data plane programming with P4: Fundamentals, advances, and applied research</title>
    <author initials="F." surname="Hauser" fullname="F. Hauser">
      <organization></organization>
    </author>
    <author initials="M." surname="Häberle" fullname="M. Häberle">
      <organization></organization>
    </author>
    <author initials="D." surname="Merling" fullname="D. Merling">
      <organization></organization>
    </author>
    <author initials="S." surname="Lindner" fullname="S. Lindner">
      <organization></organization>
    </author>
    <author initials="V." surname="Gurevich" fullname="V. Gurevich">
      <organization></organization>
    </author>
    <author initials="F." surname="Zeiger" fullname="F. Zeiger">
      <organization></organization>
    </author>
    <author initials="R." surname="Frank" fullname="R. Frank">
      <organization></organization>
    </author>
    <author initials="M." surname="Menth" fullname="M. Menth">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.jnca.2022.103561"/>
<annotation>Elsevier Journal of Network and Computer Applications, vol. 212</annotation></reference>
<reference anchor="Sha22" >
  <front>
    <title>A comparative study on online machine learning techniques for network traffic streams analysis</title>
    <author initials="A." surname="Shahraki" fullname="A. Shahraki">
      <organization></organization>
    </author>
    <author initials="M." surname="Abbasi" fullname="M. Abbasi">
      <organization></organization>
    </author>
    <author initials="A." surname="Taherkordi" fullname="A. Taherkordi">
      <organization></organization>
    </author>
    <author initials="A. D." surname="Jurcut" fullname="A. D. Jurcut">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comnet.2022.108836"/>
<annotation>Elsevier Computer Networks, vol. 207</annotation></reference>
<reference anchor="Arz21" >
  <front>
    <title>Interpretable Feedback for AutoML and a Proposal for Domain-customized AutoML for Networking</title>
    <author initials="B." surname="Arzani" fullname="B. Arzani">
      <organization></organization>
    </author>
    <author initials="K." surname="Hsieh" fullname="K. Hsieh">
      <organization></organization>
    </author>
    <author initials="H." surname="Chen" fullname="H. Chen">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3484266.3487373"/>
<annotation>ACM Workshop on Hot Topics in Networks (HotNets)</annotation></reference>
<reference anchor="Sah23" >
  <front>
    <title>Deep Reinforcement Learning Approaches to Network Slice Scaling and Placement: A Survey</title>
    <author initials="N." surname="Saha" fullname="N. Saha">
      <organization></organization>
    </author>
    <author initials="M." surname="Zangooei" fullname="M. Zangooei">
      <organization></organization>
    </author>
    <author initials="M." surname="Golkarifard" fullname="M. Golkarifard">
      <organization></organization>
    </author>
    <author initials="R." surname="Boutaba" fullname="R. Boutaba">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/MCOM.006.2200534"/>
<annotation>IEEE Communications Magazine, 61(2), pp. 82-87</annotation></reference>
<reference anchor="Urb23" >
  <front>
    <title>STREAMLINE: A Simple, Transparent, End-To-End Automated Machine Learning Pipeline Facilitating Data Analysis and Algorithm Comparison</title>
    <author initials="R." surname="Urbanowicz" fullname="R. Urbanowicz">
      <organization></organization>
    </author>
    <author initials="R." surname="Zhang" fullname="R. Zhang">
      <organization></organization>
    </author>
    <author initials="Y." surname="Cui" fullname="Y. Cui">
      <organization></organization>
    </author>
    <author initials="P." surname="Suri" fullname="P. Suri">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/978-981-19-8460-0_9"/>
<annotation>Genetic Programming Theory and Practice XIX. Genetic and Evolutionary Computation. Springer</annotation></reference>
<reference anchor="Ham23" >
  <front>
    <title>Advancing aspect-based sentiment analysis with a novel architecture combining deep learning models CNN and bi-RNN with the machine learning model SVM</title>
    <author initials="S." surname="Hammi" fullname="S. Hammi">
      <organization></organization>
    </author>
    <author initials="S. M." surname="Hammami" fullname="S. M. Hammami">
      <organization></organization>
    </author>
    <author initials="L.H." surname="Belguith" fullname="L.H. Belguith">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s13278-023-01126-4"/>
<annotation>Social Network Analysis and Mining, 13, article 117</annotation></reference>
<reference anchor="Das23" >
  <front>
    <title>Ensemble Learning Techniques and Their Applications: An Overview</title>
    <author initials="A.K." surname="Dasari" fullname="A.K. Dasari">
      <organization></organization>
    </author>
    <author initials="S. K." surname="Biswas" fullname="S. K. Biswas">
      <organization></organization>
    </author>
    <author initials="D. M." surname="Thounaojam" fullname="D. M. Thounaojam">
      <organization></organization>
    </author>
    <author initials="D." surname="Devi" fullname="D. Devi">
      <organization></organization>
    </author>
    <author initials="B." surname="Purkayastha" fullname="B. Purkayastha">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/978-981-19-8086-2_85"/>
<annotation>Advances in Cognitive Science and Communications, Springer, pp. 897-912</annotation></reference>
<reference anchor="Wu24" >
  <front>
    <title>NetLLM: Adapting Large Language Models for Networking</title>
    <author initials="D." surname="Wu" fullname="D. Wu">
      <organization></organization>
    </author>
    <author initials="X." surname="Wang" fullname="X. Wang">
      <organization></organization>
    </author>
    <author initials="Y." surname="Qiao" fullname="Y. Qiao">
      <organization></organization>
    </author>
    <author initials="Z." surname="Wang" fullname="Z. Wang">
      <organization></organization>
    </author>
    <author initials="J." surname="Jiang" fullname="J. Jiang">
      <organization></organization>
    </author>
    <author initials="S." surname="Cui" fullname="S. Cui">
      <organization></organization>
    </author>
    <author initials="F." surname="Wang" fullname="F. Wang">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3651890.3672268"/>
<annotation>ACM SIGCOMM 2024 Conference</annotation></reference>
<reference anchor="Hua25" >
  <front>
    <title>Large Language Models for Networking: Applications, Enabling Techniques, and Challenges</title>
    <author initials="Y." surname="Huang" fullname="Y. Huang">
      <organization></organization>
    </author>
    <author initials="H." surname="Du" fullname="H. Du">
      <organization></organization>
    </author>
    <author initials="X." surname="Zhang" fullname="X. Zhang">
      <organization></organization>
    </author>
    <author initials="D." surname="Niyato" fullname="D. Niyato">
      <organization></organization>
    </author>
    <author initials="J." surname="Kang" fullname="J. Kang">
      <organization></organization>
    </author>
    <author initials="Z." surname="Xiong" fullname="Z. Xiong">
      <organization></organization>
    </author>
    <author initials="S." surname="Wang" fullname="S. Wang">
      <organization></organization>
    </author>
    <author initials="T." surname="Huang" fullname="T. Huang">
      <organization></organization>
    </author>
    <date year="2025"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/MNET.2024.3435752"/>
<annotation>IEEE Network, 39(1), pp. 235-242</annotation></reference>
<reference anchor="Gre19" >
  <front>
    <title>Fashion-MNIST classification based on HOG feature descriptor using SVM</title>
    <author initials="K. V." surname="Greeshma" fullname="K. V. Greeshma">
      <organization></organization>
    </author>
    <author initials="K." surname="Sreekumar" fullname="K. Sreekumar">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
<annotation>International Journal of Innovative Technology and Exploring Engineering (IJITEE), 8(5), pp. 960-962</annotation></reference>
<reference anchor="Sen24" >
  <front>
    <title>A Survey on XAI for 5G and Beyond Security: Technical Aspects, Challenges and Research Directions</title>
    <author initials="T." surname="Senevirathna" fullname="T. Senevirathna">
      <organization></organization>
    </author>
    <author initials="V. H." surname="La" fullname="V. H. La">
      <organization></organization>
    </author>
    <author initials="S." surname="Marchal" fullname="S. Marchal">
      <organization></organization>
    </author>
    <author initials="B." surname="Siniarski" fullname="B. Siniarski">
      <organization></organization>
    </author>
    <author initials="M." surname="Liyanage" fullname="M. Liyanage">
      <organization></organization>
    </author>
    <author initials="S." surname="Wang" fullname="S. Wang">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/COMST.2024.3437248"/>
<annotation>IEEE Communications Surveys &amp; Tutorials, 27(2), pp. 941-973</annotation></reference>
<reference anchor="Liu21" >
  <front>
    <title>Constraint-Aware Deep Reinforcement Learning for End-to-End Resource Orchestration in Mobile Networks</title>
    <author initials="Q." surname="Liu" fullname="Q. Liu">
      <organization></organization>
    </author>
    <author initials="N." surname="Choi" fullname="N. Choi">
      <organization></organization>
    </author>
    <author initials="T." surname="Han" fullname="T. Han">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/ICNP52444.2021.9651934"/>
<annotation>IEEE International Conference on Network Protocols (ICNP)</annotation></reference>
<reference anchor="Liu22" >
  <front>
    <title>Error Prevalence in NIDS datasets: A Case Study on CIC-IDS-2017 and CSE-CIC-IDS-2018</title>
    <author initials="L." surname="Liu" fullname="L. Liu">
      <organization></organization>
    </author>
    <author initials="G." surname="Engelen" fullname="G. Engelen">
      <organization></organization>
    </author>
    <author initials="T." surname="Lynar" fullname="T. Lynar">
      <organization></organization>
    </author>
    <author initials="D." surname="Essam" fullname="D. Essam">
      <organization></organization>
    </author>
    <author initials="W." surname="Joosen" fullname="W. Joosen">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/CNS56114.2022.9947235"/>
<annotation>IEEE Conference on Communications and Network Security (CNS)</annotation></reference>
<reference anchor="Hoa21" >
  <front>
    <title>Domain name encryption is not enough: Privacy leakage via IP-based website fingerprinting</title>
    <author initials="N. P." surname="Hoang" fullname="N. P. Hoang">
      <organization></organization>
    </author>
    <author initials="A. A." surname="Niaki" fullname="A. A. Niaki">
      <organization></organization>
    </author>
    <author initials="P." surname="Gill" fullname="P. Gill">
      <organization></organization>
    </author>
    <author initials="M." surname="Polychronakis" fullname="M. Polychronakis">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
<annotation>Proceedings on Privacy Enhancing Technologies, 2021(4), pp. 420-440</annotation></reference>
<reference anchor="Liu21b" >
  <front>
    <title>When Machine Learning Meets Privacy: A Survey and Outlook</title>
    <author initials="B." surname="Liu" fullname="B. Liu">
      <organization></organization>
    </author>
    <author initials="M." surname="Ding" fullname="M. Ding">
      <organization></organization>
    </author>
    <author initials="S." surname="Shaham" fullname="S. Shaham">
      <organization></organization>
    </author>
    <author initials="W." surname="Rahayu" fullname="W. Rahayu">
      <organization></organization>
    </author>
    <author initials="F." surname="Farokhi" fullname="F. Farokhi">
      <organization></organization>
    </author>
    <author initials="Z." surname="Lin" fullname="Z. Lin">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3436755"/>
<annotation>ACM Computing Surveys, 54(2)</annotation></reference>
<reference anchor="Jie22" >
  <front>
    <title>Graph-based deep learning for communication networks: A survey</title>
    <author initials="W." surname="Jiang" fullname="W. Jiang">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comcom.2021.12.015"/>
<annotation>Computer Communications, Elsevier, vol. 185, pp. 40-54</annotation></reference>
<reference anchor="Sha21" >
  <front>
    <title>FlowPic: A Generic Representation for Encrypted Traffic Classification and Applications Identification</title>
    <author initials="T." surname="Shapira" fullname="T. Shapira">
      <organization></organization>
    </author>
    <author initials="Y." surname="Shavitt" fullname="Y. Shavitt">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TNSM.2021.3071441"/>
<annotation>IEEE Transactions on Network and Service Management, 18(2), pp. 1218-1232</annotation></reference>
<reference anchor="Sul23" >
  <front>
    <title>Coordinated Slicing and Admission Control Using Multi-Agent Deep Reinforcement Learning</title>
    <author initials="M." surname="Sulaiman" fullname="M. Sulaiman">
      <organization></organization>
    </author>
    <author initials="A." surname="Moayyedi" fullname="A. Moayyedi">
      <organization></organization>
    </author>
    <author initials="M." surname="Ahmadi" fullname="M. Ahmadi">
      <organization></organization>
    </author>
    <author initials="M. A." surname="Salahuddin" fullname="M. A. Salahuddin">
      <organization></organization>
    </author>
    <author initials="R." surname="Boutaba" fullname="R. Boutaba">
      <organization></organization>
    </author>
    <author initials="A." surname="Saleh" fullname="A. Saleh">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TNSM.2022.3222589"/>
<annotation>IEEE Transactions on Network and Service Management, 20(2), pp. 1110-1124</annotation></reference>
<reference anchor="Vin21" >
  <front>
    <title>Timely admission control for network slicing in 5G with machine learning</title>
    <author initials="M." surname="Vincenzi" fullname="M. Vincenzi">
      <organization></organization>
    </author>
    <author initials="E." surname="Lopez-Aguilera" fullname="E. Lopez-Aguilera">
      <organization></organization>
    </author>
    <author initials="E." surname="Garcia-Villegas" fullname="E. Garcia-Villegas">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/ACCESS.2021.3111143"/>
<annotation>IEEE Access, vol. 9, pp. 127595-127610</annotation></reference>
<reference anchor="Att23" >
  <front>
    <title>VNF and CNF Placement in 5G: Recent Advances and Future Trends</title>
    <author initials="W." surname="Attaoui" fullname="W. Attaoui">
      <organization></organization>
    </author>
    <author initials="E." surname="Sabir" fullname="E. Sabir">
      <organization></organization>
    </author>
    <author initials="H." surname="Elbiaze" fullname="H. Elbiaze">
      <organization></organization>
    </author>
    <author initials="M." surname="Guizani" fullname="M. Guizani">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TNSM.2023.3264005"/>
<annotation>IEEE Transactions on Network and Service Management, 20(4), pp. 4698-4733</annotation></reference>
<reference anchor="Col22" >
  <front>
    <title>LossLeaP: Learning to Predict for Intent-Based Networking</title>
    <author initials="A." surname="Collet" fullname="A. Collet">
      <organization></organization>
    </author>
    <author initials="A." surname="Banchs" fullname="A. Banchs">
      <organization></organization>
    </author>
    <author initials="M." surname="Fiore" fullname="M. Fiore">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/INFOCOM48880.2022.9796918"/>
<annotation>IEEE Conference on Computer Communications (INFOCOM)</annotation></reference>
<reference anchor="Ami24" >
  <front>
    <title>Using Machine Learning to Improve Interactive Visualizations for Large Collected Traffic Detector Data</title>
    <author initials="R. M." surname="Amin" fullname="R. M. Amin">
      <organization></organization>
    </author>
    <author initials="P." surname="Hammer" fullname="P. Hammer">
      <organization></organization>
    </author>
    <author initials="A." surname="Butz" fullname="A. Butz">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3640543.3645177"/>
<annotation>Proceedings of the 29th International Conference on Intelligent User Interfaces (IUI'24), ACM, pp. 803-816</annotation></reference>
<reference anchor="Zha21" >
  <front>
    <title>Network traffic classification for data fusion: A survey</title>
    <author initials="J." surname="Zhao" fullname="J. Zhao">
      <organization></organization>
    </author>
    <author initials="X." surname="Jing" fullname="X. Jing">
      <organization></organization>
    </author>
    <author initials="Z." surname="Yan" fullname="Z. Yan">
      <organization></organization>
    </author>
    <author initials="W." surname="Pedrycz" fullname="W. Pedrycz">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.inffus.2021.02.009"/>
<annotation>Information Fusion, Elsevier, vol. 72, pp. 22-47</annotation></reference>
<reference anchor="Mut24" >
  <front>
    <title>Digital twin (DT)-based predictive maintenance of a 6G communication network</title>
    <author initials="F." surname="Muteba" fullname="F. Muteba">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.procs.2024.06.058"/>
<annotation>Procedia Computer Science, Elsevier, vol. 238, pp. 544-549</annotation></reference>
<reference anchor="Gin24" >
  <front>
    <title>TrafficGPT: An LLM Approach for Open-Set Encrypted Traffic Classification</title>
    <author initials="Y." surname="Ginige" fullname="Y. Ginige">
      <organization></organization>
    </author>
    <author initials="T." surname="Dahanayaka" fullname="T. Dahanayaka">
      <organization></organization>
    </author>
    <author initials="S." surname="Seneviratne" fullname="S. Seneviratne">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3674213.3674217"/>
<annotation>Asian Internet Engineering Conference (AINTEC), pp. 26-35</annotation></reference>
<reference anchor="Ake24" >
  <front>
    <title>Encrypted Traffic Classification at Line Rate in Programmable Switches with Machine Learning</title>
    <author initials="A. T. J." surname="Akem" fullname="A. T. J. Akem">
      <organization></organization>
    </author>
    <author initials="G." surname="Fraysse" fullname="G. Fraysse">
      <organization></organization>
    </author>
    <author initials="M." surname="Fiore" fullname="M. Fiore">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/NOMS59830.2024.10575394"/>
<annotation>IEEE/IFIP Network Operations and Management Symposium (NOMS)</annotation></reference>
<reference anchor="Sin22" >
  <front>
    <title>Optimization of reinforcement routing for wireless mesh network using machine learning and high-performance computing</title>
    <author initials="A." surname="Singh" fullname="A. Singh">
      <organization></organization>
    </author>
    <author initials="S." surname="Prakash" fullname="S. Prakash">
      <organization></organization>
    </author>
    <author initials="S." surname="Singh" fullname="S. Singh">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1002/cpe.6960"/>
<annotation>Concurrency and Computation: Practice and Experience, 34(15)</annotation></reference>
<reference anchor="Wan24" >
  <front>
    <title>Combine intra- and inter-flow: A multimodal encrypted traffic classification model driven by diverse features</title>
    <author initials="X." surname="Wang" fullname="X. Wang">
      <organization></organization>
    </author>
    <author initials="Q." surname="Yuan" fullname="Q. Yuan">
      <organization></organization>
    </author>
    <author initials="Y." surname="Wang" fullname="Y. Wang">
      <organization></organization>
    </author>
    <author initials="G." surname="Gou" fullname="G. Gou">
      <organization></organization>
    </author>
    <author initials="C." surname="Gu" fullname="C. Gu">
      <organization></organization>
    </author>
    <author initials="G." surname="Yu" fullname="G. Yu">
      <organization></organization>
    </author>
    <author initials="G." surname="Xiong" fullname="G. Xiong">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comnet.2024.110403"/>
<annotation>Computer Networks, Elsevier, vol. 245</annotation></reference>
<reference anchor="Naj24" >
  <front>
    <title>Enhanced Network Traffic Classification with Machine Learning Algorithms</title>
    <author initials="I. A." surname="Najm" fullname="I. A. Najm">
      <organization></organization>
    </author>
    <author initials="A. H." surname="Saeed" fullname="A. H. Saeed">
      <organization></organization>
    </author>
    <author initials="B.A." surname="Ahmad" fullname="B.A. Ahmad">
      <organization></organization>
    </author>
    <author initials="S. R." surname="Ahmed" fullname="S. R. Ahmed">
      <organization></organization>
    </author>
    <author initials="R." surname="Sekhar" fullname="R. Sekhar">
      <organization></organization>
    </author>
    <author initials="P." surname="Shah" fullname="P. Shah">
      <organization></organization>
    </author>
    <author initials="B.S." surname="Veena" fullname="B.S. Veena">
      <organization></organization>
    </author>
    <date year="2024"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3660853.3660935"/>
<annotation>Proceedings of the Cognitive Models and Artificial Intelligence Conference, pp. 322-327</annotation></reference>
<reference anchor="Sha25" >
  <front>
    <title>Chapter 26 - Future now: unleashing the potentials of machine learning algorithms in 6G</title>
    <author initials="G." surname="Sharma" fullname="G. Sharma">
      <organization></organization>
    </author>
    <author initials="B." surname="Priya" fullname="B. Priya">
      <organization></organization>
    </author>
    <date year="2025"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/B978-0-44-330012-7.00034-5"/>
<annotation>Advances in Computational Methods and Modeling for Science and Engineering, Morgan Kaufmann, pp. 331-339</annotation></reference>
<reference anchor="Tan21" >
  <front>
    <title>Comprehensive Survey on Machine Learning in Vehicular Network: Technology, Applications and Challenges</title>
    <author initials="F." surname="Tang" fullname="F. Tang">
      <organization></organization>
    </author>
    <author initials="B." surname="Mao" fullname="B. Mao">
      <organization></organization>
    </author>
    <author initials="N." surname="Kato" fullname="N. Kato">
      <organization></organization>
    </author>
    <author initials="G." surname="Gui" fullname="G. Gui">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/COMST.2021.3089688"/>
<annotation>IEEE Communications Surveys &amp; Tutorials, 23(3), pp. 2027-2057</annotation></reference>
<reference anchor="Ke23" >
  <front>
    <title>Adversarial Machine Learning for Network Intrusion Detection Systems: A Comprehensive Survey</title>
    <author initials="K." surname="He" fullname="K. He">
      <organization></organization>
    </author>
    <author initials="D. D." surname="Kim" fullname="D. D. Kim">
      <organization></organization>
    </author>
    <author initials="M. R." surname="Asghar" fullname="M. R. Asghar">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/COMST.2022.3233793"/>
<annotation>IEEE Communications Surveys &amp; Tutorials, 25(1), pp. 538-566</annotation></reference>
<reference anchor="Mcc93" >
  <front>
    <title>The BSD packet filter: a new architecture for user-level packet capture</title>
    <author initials="S." surname="McCanne" fullname="Steve McCanne">
      <organization></organization>
    </author>
    <author initials="V." surname="Jacobson" fullname="Van Jacobson">
      <organization></organization>
    </author>
    <date year="1993"/>
  </front>
<annotation>USENIX Winter Conference</annotation></reference>
<reference anchor="Ste92" >
  <front>
    <title>A Diagnosis Expert System for Network Traffic Management</title>
    <author initials="D." surname="Stern" fullname="Daniel Stern">
      <organization></organization>
    </author>
    <author initials="P." surname="Chemouil" fullname="Prosper Chemouil">
      <organization></organization>
    </author>
    <date year="1992"/>
  </front>
<annotation>Networks, Kobe, Japan</annotation></reference>
<reference anchor="Abd10" >
  <front>
    <title>Comparison of Machine Learning Algorithms Performance in Detecting Network Intrusion</title>
    <author initials="K. A." surname="Jalil" fullname="Kamarularifin Abd Jalil">
      <organization></organization>
    </author>
    <author initials="M. H." surname="Kamarudin" fullname="Muhammad Hilmi Kamarudin">
      <organization></organization>
    </author>
    <author initials="M. N." surname="Masrek" fullname="Mohamad Noorman Masrek">
      <organization></organization>
    </author>
    <date year="2010"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/ICNIT.2010.5508526"/>
<annotation>IEEE International Conference on Networking and Information Technology (ICNIT), pp. 221-226</annotation></reference>
<reference anchor="Beg19" >
  <front>
    <title>DeepCog: Cognitive Network Management in Sliced 5G Networks with Deep Learning</title>
    <author initials="D." surname="Bega" fullname="D. Bega">
      <organization></organization>
    </author>
    <author initials="M." surname="Gramaglia" fullname="M. Gramaglia">
      <organization></organization>
    </author>
    <author initials="M." surname="Fiore" fullname="M. Fiore">
      <organization></organization>
    </author>
    <author initials="A." surname="Banchs" fullname="A. Banchs">
      <organization></organization>
    </author>
    <author initials="X." surname="Costa-Perez" fullname="X. Costa-Perez">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/INFOCOM.2019.8737488"/>
<annotation>IEEE INFOCOM</annotation></reference>
<reference anchor="Bos13" >
  <front>
    <title>Forwarding metamorphosis: Fast programmable match-action processing in hardware for SDN</title>
    <author initials="P." surname="Bosshart" fullname="Pat Bosshart">
      <organization></organization>
    </author>
    <author initials="G." surname="Gibb" fullname="Glen Gibb">
      <organization></organization>
    </author>
    <author initials="H.-S." surname="Kim" fullname="Hun-Seok Kim">
      <organization></organization>
    </author>
    <author initials="G." surname="Varghese" fullname="George Varghese">
      <organization></organization>
    </author>
    <author initials="N." surname="McKeown" fullname="Nick McKeown">
      <organization></organization>
    </author>
    <author initials="M." surname="Izzard" fullname="Martin Izzard">
      <organization></organization>
    </author>
    <author initials="F." surname="Mujica" fullname="Fernando Mujica">
      <organization></organization>
    </author>
    <author initials="M." surname="Horowitz" fullname="Mark Horowitz">
      <organization></organization>
    </author>
    <date year="2013"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/2486001.2486011"/>
<annotation>ACM SIGCOMM</annotation></reference>
<reference anchor="Bos14" >
  <front>
    <title>P4: Programming Protocol-Independent Packet Processors</title>
    <author initials="P." surname="Bosshart" fullname="Pat Bosshart">
      <organization></organization>
    </author>
    <author initials="D." surname="Daly" fullname="Dan Daly">
      <organization></organization>
    </author>
    <author initials="G." surname="Gibb" fullname="Glen Gibb">
      <organization></organization>
    </author>
    <author initials="M." surname="Izzard" fullname="Martin Izzard">
      <organization></organization>
    </author>
    <author initials="N." surname="McKeown" fullname="Nick McKeown">
      <organization></organization>
    </author>
    <author initials="J." surname="Rexford" fullname="Jennifer Rexford">
      <organization></organization>
    </author>
    <author initials="C." surname="Schlesinger" fullname="Cole Schlesinger">
      <organization></organization>
    </author>
    <author initials="D." surname="Talayco" fullname="Dan Talayco">
      <organization></organization>
    </author>
    <author initials="A." surname="Vahdat" fullname="Amin Vahdat">
      <organization></organization>
    </author>
    <author initials="G." surname="Varghese" fullname="George Varghese">
      <organization></organization>
    </author>
    <author initials="D." surname="Walker" fullname="David Walker">
      <organization></organization>
    </author>
    <date year="2014"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/2656877.2656890"/>
<annotation>ACM SIGCOMM Computer Communication Review, 44(3), pp. 87-95</annotation></reference>
<reference anchor="Dat18" >
  <front>
    <title>P4Guard: Designing P4 Based Firewall</title>
    <author initials="R." surname="Datta" fullname="R. Datta">
      <organization></organization>
    </author>
    <author initials="S." surname="Choi" fullname="S. Choi">
      <organization></organization>
    </author>
    <author initials="A." surname="Chowdhary" fullname="A. Chowdhary">
      <organization></organization>
    </author>
    <author initials="Y." surname="Park" fullname="Y. Park">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/MILCOM.2018.8599726"/>
<annotation>IEEE Military Communications Conference (MILCOM)</annotation></reference>
<reference anchor="Che19" >
  <front>
    <title>P4-Enabled Bandwidth Management</title>
    <author initials="Y." surname="Chen" fullname="Y. Chen">
      <organization></organization>
    </author>
    <author initials="L." surname="Yen" fullname="L. Yen">
      <organization></organization>
    </author>
    <author initials="W." surname="Wang" fullname="W. Wang">
      <organization></organization>
    </author>
    <author initials="C." surname="Chuang" fullname="C. Chuang">
      <organization></organization>
    </author>
    <author initials="Y." surname="Liu" fullname="Y. Liu">
      <organization></organization>
    </author>
    <author initials="C." surname="Tseng" fullname="C. Tseng">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.23919/APNOMS.2019.8892909"/>
<annotation>Asia-Pacific Network Operations and Management Symposium (APNOMS)</annotation></reference>
<reference anchor="Cha18" >
  <front>
    <title>Oko: Extending Open vSwitch with Stateful Filters</title>
    <author initials="P." surname="Chaignon" fullname="Paul Chaignon">
      <organization></organization>
    </author>
    <author initials="K." surname="Lazri" fullname="Kahina Lazri">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="T." surname="Delmas" fullname="Thibault Delmas">
      <organization></organization>
    </author>
    <author initials="O." surname="Festor" fullname="Olivier Festor">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3185467.3185496"/>
<annotation>ACM Symposium on SDN Research (SOSR)</annotation></reference>
<reference anchor="czb20" >
  <front>
    <title>Network Management 2030: Operations and Control of Network 2030 Services</title>
    <author initials="A." surname="Clemm" fullname="A. Clemm">
      <organization></organization>
    </author>
    <author initials="M. F." surname="Zhani" fullname="M. F. Zhani">
      <organization></organization>
    </author>
    <author initials="R." surname="Boutaba" fullname="R. Boutaba">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s10922-020-09517-0"/>
<annotation>Springer Journal of Network and Systems Management (JNSM), 28(4), pp. 721-750</annotation></reference>
<reference anchor="Hua19" >
  <front>
    <title>Addressing the Loss-Metric Mismatch with Adaptive Loss Alignment</title>
    <author initials="C." surname="Huang" fullname="Chen Huang">
      <organization></organization>
    </author>
    <author initials="S." surname="Zhai" fullname="Shuangfei Zhai">
      <organization></organization>
    </author>
    <author initials="W." surname="Talbott" fullname="Walter Talbott">
      <organization></organization>
    </author>
    <author initials="M. A." surname="Bautista" fullname="Miguel Angel Bautista">
      <organization></organization>
    </author>
    <author initials="S.-Y." surname="Sun" fullname="Shih-Yu Sun">
      <organization></organization>
    </author>
    <author initials="C." surname="Guestrin" fullname="Carlos Guestrin">
      <organization></organization>
    </author>
    <author initials="J." surname="Susskind" fullname="Josh Susskind">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
<annotation>International Conference on Machine Learning (ICML), PMLR 97</annotation></reference>
<reference anchor="Jos22" >
  <front>
    <title>NetREC: Network-wide in-network REal-value Computation</title>
    <author initials="M." surname="Jose" fullname="Matthews Jose">
      <organization></organization>
    </author>
    <author initials="K." surname="Lazri" fullname="Kahina Lazri">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="O." surname="Festor" fullname="Olivier Festor">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/NetSoft54395.2022.9844118"/>
<annotation>IEEE International Conference on Network Softwarization (NetSoft)</annotation></reference>
<reference anchor="Lee21" >
  <front>
    <title>PEBBLE: Feedback-Efficient Interactive Reinforcement Learning via Relabeling Experience and Unsupervised Pre-training</title>
    <author initials="K." surname="Lee" fullname="K. Lee">
      <organization></organization>
    </author>
    <author initials="L." surname="Smith" fullname="L. Smith">
      <organization></organization>
    </author>
    <author initials="P." surname="Abbeel" fullname="P. Abbeel">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
<annotation>International Conference on Machine Learning (ICML), PMLR 139</annotation></reference>
<reference anchor="Liu16" >
  <front>
    <title>One Sketch to Rule Them All: Rethinking Network Flow Monitoring with UnivMon</title>
    <author initials="Z." surname="Liu" fullname="Zaoxing Liu">
      <organization></organization>
    </author>
    <author initials="A." surname="Manousis" fullname="Antonis Manousis">
      <organization></organization>
    </author>
    <author initials="G." surname="Vorsanger" fullname="Gregory Vorsanger">
      <organization></organization>
    </author>
    <author initials="V." surname="Sekar" fullname="Vyas Sekar">
      <organization></organization>
    </author>
    <author initials="V." surname="Braverman" fullname="Vladimir Braverman">
      <organization></organization>
    </author>
    <date year="2016"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/2934872.2934906"/>
<annotation>ACM SIGCOMM Conference</annotation></reference>
<reference anchor="Yan18" >
  <front>
    <title>Elastic sketch: adaptive and fast network-wide measurements</title>
    <author initials="T." surname="Yang" fullname="Tong Yang">
      <organization></organization>
    </author>
    <author initials="J." surname="Jiang" fullname="Jie Jiang">
      <organization></organization>
    </author>
    <author initials="P." surname="Liu" fullname="Peng Liu">
      <organization></organization>
    </author>
    <author initials="Q." surname="Huang" fullname="Qun Huang">
      <organization></organization>
    </author>
    <author initials="J." surname="Gong" fullname="Junzhi Gong">
      <organization></organization>
    </author>
    <author initials="Y." surname="Zhou" fullname="Yang Zhou">
      <organization></organization>
    </author>
    <author initials="R." surname="Miao" fullname="Rui Miao">
      <organization></organization>
    </author>
    <author initials="X." surname="Li" fullname="Xiaoming Li">
      <organization></organization>
    </author>
    <author initials="S." surname="Uhlig" fullname="Steve Uhlig">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3230543.3230544"/>
<annotation>ACM SIGCOMM Conference</annotation></reference>
<reference anchor="Gup18" >
  <front>
    <title>Sonata: query-driven streaming network telemetry</title>
    <author initials="A." surname="Gupta" fullname="Arpit Gupta">
      <organization></organization>
    </author>
    <author initials="R." surname="Harrison" fullname="Rob Harrison">
      <organization></organization>
    </author>
    <author initials="M." surname="Canini" fullname="Marco Canini">
      <organization></organization>
    </author>
    <author initials="N." surname="Feamster" fullname="Nick Feamster">
      <organization></organization>
    </author>
    <author initials="J." surname="Rexford" fullname="Jennifer Rexford">
      <organization></organization>
    </author>
    <author initials="W." surname="Willinger" fullname="Walter Willinger">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3230543.3230555"/>
<annotation>ACM SIGCOMM Conference</annotation></reference>
<reference anchor="Mar18" >
  <front>
    <title>Exploiting External Events for Resource Adaptation in Virtual Computer and Network Systems</title>
    <author initials="P." surname="Martinez-Julia" fullname="Pedro Martinez-Julia">
      <organization></organization>
    </author>
    <author initials="V. P." surname="Kafle" fullname="Ved P. Kafle">
      <organization></organization>
    </author>
    <author initials="H." surname="Harai" fullname="Hiroaki Harai">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TNSM.2018.2794530"/>
<annotation>IEEE Transactions on Network and Service Management, 15(2), pp. 555-566</annotation></reference>
<reference anchor="Mar20" >
  <front>
    <title>Explained Intelligent Management Decisions in Virtual Networks and Network Slices</title>
    <author initials="P." surname="Martinez-Julia" fullname="Pedro Martinez-Julia">
      <organization></organization>
    </author>
    <author initials="V. P." surname="Kafle" fullname="Ved P. Kafle">
      <organization></organization>
    </author>
    <author initials="H." surname="Asaeda" fullname="H. Asaeda">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/ICIN48450.2020.9059462"/>
<annotation>Conference on Innovation in Clouds, Internet and Networks and Workshops (ICIN)</annotation></reference>
<reference anchor="Ngu20" >
  <front>
    <title>Efficient SDN-based traffic monitoring in IoT networks with double deep Q-network</title>
    <author initials="T. G." surname="Nguyen" fullname="T. G. Nguyen">
      <organization></organization>
    </author>
    <author initials="T. V." surname="Phan" fullname="T. V. Phan">
      <organization></organization>
    </author>
    <author initials="D. T." surname="Hoang" fullname="D. T. Hoang">
      <organization></organization>
    </author>
    <author initials="T. N." surname="Nguyen" fullname="T. N. Nguyen">
      <organization></organization>
    </author>
    <author initials="C." surname="So-In" fullname="C. So-In">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/978-3-030-66046-8_3"/>
<annotation>International Conference on Computational Data and Social Networks (CSoNet), Springer LNCS</annotation></reference>
<reference anchor="Tan20" >
  <front>
    <title>Classification-assisted Query Processing for Network Telemetry</title>
    <author initials="G." surname="Tangari" fullname="G. Tangari">
      <organization></organization>
    </author>
    <author initials="M." surname="Charalambides" fullname="M. Charalambides">
      <organization></organization>
    </author>
    <author initials="G." surname="Pavlou" fullname="G. Pavlou">
      <organization></organization>
    </author>
    <author initials="C." surname="Grazian" fullname="C. Grazian">
      <organization></organization>
    </author>
    <author initials="D." surname="Tuncer" fullname="D. Tuncer">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
<annotation>Network Traffic Measurement and Analysis Conference (TMA)</annotation></reference>
<reference anchor="Tan21b" >
  <front>
    <title>In-band Network Telemetry: A Survey</title>
    <author initials="L." surname="Tan" fullname="L. Tan">
      <organization></organization>
    </author>
    <author initials="W." surname="Su" fullname="W. Su">
      <organization></organization>
    </author>
    <author initials="W." surname="Zhang" fullname="W. Zhang">
      <organization></organization>
    </author>
    <author initials="J." surname="Lv" fullname="J. Lv">
      <organization></organization>
    </author>
    <author initials="Z." surname="Zhang" fullname="Z. Zhang">
      <organization></organization>
    </author>
    <author initials="J." surname="Miao" fullname="J. Miao">
      <organization></organization>
    </author>
    <author initials="X." surname="Liu" fullname="X. Liu">
      <organization></organization>
    </author>
    <author initials="N." surname="Li" fullname="N. Li">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comnet.2020.107763"/>
<annotation>Computer Networks, Elsevier, vol. 186</annotation></reference>
<reference anchor="Val17" >
  <front>
    <title>Learning to route</title>
    <author initials="A." surname="Valadarsky" fullname="A. Valadarsky">
      <organization></organization>
    </author>
    <author initials="M." surname="Schapira" fullname="M. Schapira">
      <organization></organization>
    </author>
    <author initials="D." surname="Shahaf" fullname="D. Shahaf">
      <organization></organization>
    </author>
    <author initials="A." surname="Tamar" fullname="A. Tamar">
      <organization></organization>
    </author>
    <date year="2017"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3152434.3152441"/>
<annotation>ACM Workshop on Hot Topics in Networks (HotNets)</annotation></reference>
<reference anchor="Xu18" >
  <front>
    <title>Experience-driven networking: A deep reinforcement learning based approach</title>
    <author initials="Z." surname="Xu" fullname="Z. Xu">
      <organization></organization>
    </author>
    <author initials="J." surname="Tang" fullname="J. Tang">
      <organization></organization>
    </author>
    <author initials="J." surname="Meng" fullname="J. Meng">
      <organization></organization>
    </author>
    <author initials="W." surname="Zhang" fullname="W. Zhang">
      <organization></organization>
    </author>
    <author initials="Y." surname="Wang" fullname="Y. Wang">
      <organization></organization>
    </author>
    <author initials="C. H." surname="Liu" fullname="C. H. Liu">
      <organization></organization>
    </author>
    <author initials="D." surname="Yang" fullname="D. Yang">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/INFOCOM.2018.8485853"/>
<annotation>IEEE INFOCOM</annotation></reference>
<reference anchor="tl1" >
  <front>
    <title>Transfer learning</title>
    <author initials="L." surname="Torrey" fullname="Lisa Torrey">
      <organization></organization>
    </author>
    <author initials="J." surname="Shavlik" fullname="Jude Shavlik">
      <organization></organization>
    </author>
    <date year="2010"/>
  </front>
  <seriesInfo name="DOI" value="10.4018/978-1-60566-766-9.ch011"/>
<annotation>Handbook of research on machine learning applications and trends: algorithms, methods, and techniques, IGI Global, pp. 242-264</annotation></reference>
<reference anchor="gnn1" >
  <front>
    <title>Relational inductive biases, deep learning, and graph networks</title>
    <author initials="P. W." surname="Battaglia" fullname="Peter W. Battaglia">
      <organization></organization>
    </author>
    <author initials="et" surname="al." fullname="et al.">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.48550/arXiv.1806.01261"/>
<annotation>arXiv preprint arXiv:1806.01261</annotation></reference>
<reference anchor="gnn2" >
  <front>
    <title>Unveiling the potential of Graph Neural Networks for network modeling and optimization in SDN</title>
    <author initials="K." surname="Rusek" fullname="K. Rusek">
      <organization></organization>
    </author>
    <author initials="J." surname="Suárez-Varela" fullname="J. Suárez-Varela">
      <organization></organization>
    </author>
    <author initials="A." surname="Mestres" fullname="A. Mestres">
      <organization></organization>
    </author>
    <author initials="P." surname="Barlet-Ros" fullname="P. Barlet-Ros">
      <organization></organization>
    </author>
    <author initials="A." surname="Cabellos-Aparicio" fullname="A. Cabellos-Aparicio">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3314148.3314357"/>
<annotation>ACM Symposium on SDN Research (SOSR)</annotation></reference>
<reference anchor="Rex06" >
  <front>
    <title>Route optimization in IP networks</title>
    <author initials="J." surname="Rexford" fullname="Jennifer Rexford">
      <organization></organization>
    </author>
    <date year="2006"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/978-0-387-30165-5_24"/>
<annotation>Handbook of Optimization in Telecommunications, Springer, pp. 679-700</annotation></reference>
<reference anchor="Kr15" >
  <front>
    <title>Software-defined networking: A comprehensive survey</title>
    <author initials="D." surname="Kreutz" fullname="D. Kreutz">
      <organization></organization>
    </author>
    <author initials="F. M." surname="Ramos" fullname="F. M. Ramos">
      <organization></organization>
    </author>
    <author initials="P. E." surname="Verissimo" fullname="P. E. Verissimo">
      <organization></organization>
    </author>
    <author initials="C. E." surname="Rothenberg" fullname="C. E. Rothenberg">
      <organization></organization>
    </author>
    <author initials="S." surname="Azodolmolky" fullname="S. Azodolmolky">
      <organization></organization>
    </author>
    <author initials="S." surname="Uhlig" fullname="S. Uhlig">
      <organization></organization>
    </author>
    <date year="2015"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/JPROC.2014.2371999"/>
<annotation>Proceedings of the IEEE, 103(1), pp. 14-76</annotation></reference>
<reference anchor="XAI" >
  <front>
    <title>Explainable artificial intelligence: Understanding, visualizing and interpreting deep learning models</title>
    <author initials="W." surname="Samek" fullname="Wojciech Samek">
      <organization></organization>
    </author>
    <author initials="T." surname="Wiegand" fullname="Thomas Wiegand">
      <organization></organization>
    </author>
    <author initials="K.-R." surname="Müller" fullname="Klaus-Robert Müller">
      <organization></organization>
    </author>
    <date year="2017"/>
  </front>
<annotation>ITU Journal: ICT Discoveries, Special Issue No. 1</annotation></reference>
<reference anchor="Men20" >
  <front>
    <title>Interpreting Deep Learning-Based Networking Systems</title>
    <author initials="Z." surname="Meng" fullname="Zili Meng">
      <organization></organization>
    </author>
    <author initials="M." surname="Wang" fullname="Minhu Wang">
      <organization></organization>
    </author>
    <author initials="J." surname="Bai" fullname="Jiasong Bai">
      <organization></organization>
    </author>
    <author initials="M." surname="Xu" fullname="Mingwei Xu">
      <organization></organization>
    </author>
    <author initials="H." surname="Mao" fullname="Hongzi Mao">
      <organization></organization>
    </author>
    <author initials="H." surname="Hu" fullname="Hongxin Hu">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3387514.3405859"/>
<annotation>ACM SIGCOMM</annotation></reference>
<reference anchor="Mor18" >
  <front>
    <title>Ray: A Distributed Framework for Emerging AI Applications</title>
    <author initials="P." surname="Moritz" fullname="Philipp Moritz">
      <organization></organization>
    </author>
    <author initials="R." surname="Nishihara" fullname="Robert Nishihara">
      <organization></organization>
    </author>
    <author initials="S." surname="Wang" fullname="Stephanie Wang">
      <organization></organization>
    </author>
    <author initials="A." surname="Tumanov" fullname="Alexey Tumanov">
      <organization></organization>
    </author>
    <author initials="R." surname="Liaw" fullname="Richard Liaw">
      <organization></organization>
    </author>
    <author initials="E." surname="Liang" fullname="Eric Liang">
      <organization></organization>
    </author>
    <author initials="M." surname="Elibol" fullname="Melih Elibol">
      <organization></organization>
    </author>
    <author initials="Z." surname="Yang" fullname="Zongheng Yang">
      <organization></organization>
    </author>
    <author initials="W." surname="Paul" fullname="William Paul">
      <organization></organization>
    </author>
    <author initials="M." surname="Jordan" fullname="Michael Jordan">
      <organization></organization>
    </author>
    <author initials="I." surname="Stoica" fullname="Ion Stoica">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
<annotation>USENIX Symposium on Operating Systems Design and Implementation (OSDI)</annotation></reference>
<reference anchor="Puj21" >
  <front>
    <title>NetXplain: Real-time explainability of graph neural networks applied to networking</title>
    <author initials="D." surname="Pujol-Perich" fullname="David Pujol-Perich">
      <organization></organization>
    </author>
    <author initials="J." surname="Suárez-Varela" fullname="José Suárez-Varela">
      <organization></organization>
    </author>
    <author initials="S." surname="Xiao" fullname="Shihan Xiao">
      <organization></organization>
    </author>
    <author initials="B." surname="Wu" fullname="Bo Wu">
      <organization></organization>
    </author>
    <author initials="A." surname="Cabellos-Aparicio" fullname="Albert Cabellos-Aparicio">
      <organization></organization>
    </author>
    <author initials="P." surname="Barlet-Ros" fullname="Pere Barlet-Ros">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.52953/AKZH5253"/>
<annotation>ITU Journal on Future and Evolving Technologies, 2(4), pp. 57-66</annotation></reference>
<reference anchor="Li19" >
  <front>
    <title>A Framework for Qualitative Communications Using Big Packet Protocol</title>
    <author initials="R." surname="Li" fullname="Richard Li">
      <organization></organization>
    </author>
    <author initials="K." surname="Makhijani" fullname="Kiran Makhijani">
      <organization></organization>
    </author>
    <author initials="H." surname="Yousefi" fullname="Hamed Yousefi">
      <organization></organization>
    </author>
    <author initials="C." surname="Westphal" fullname="Cedric Westphal">
      <organization></organization>
    </author>
    <author initials="L." surname="Dong" fullname="Lijun Dong">
      <organization></organization>
    </author>
    <author initials="T." surname="Wauters" fullname="Tim Wauters">
      <organization></organization>
    </author>
    <author initials="F. D." surname="Turck" fullname="Filip De Turck">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3341558.3342201"/>
<annotation>ACM SIGCOMM Workshop on Networking for Emerging Applications and Technologies (NEAT)</annotation></reference>
<reference anchor="Hir15" >
  <front>
    <title>Crowdsourced network measurements: Benefits and best practices</title>
    <author initials="M." surname="Hirth" fullname="Matthias Hirth">
      <organization></organization>
    </author>
    <author initials="T." surname="Hossfeld" fullname="Tobias Hossfeld">
      <organization></organization>
    </author>
    <author initials="M." surname="Mellia" fullname="Marco Mellia">
      <organization></organization>
    </author>
    <author initials="C." surname="Schwartz" fullname="Christian Schwartz">
      <organization></organization>
    </author>
    <author initials="F." surname="Lehrieder" fullname="Frank Lehrieder">
      <organization></organization>
    </author>
    <date year="2015"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comnet.2015.07.003"/>
<annotation>Computer Networks, vol. 90, pp. 85-98</annotation></reference>
<reference anchor="Bri19" >
  <front>
    <title>Transparent and Service-Agnostic Monitoring of Encrypted Web Traffic</title>
    <author initials="P.-O." surname="Brissaud" fullname="Pierre-Olivier Brissaud">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="I." surname="Chrisment" fullname="Isabelle Chrisment">
      <organization></organization>
    </author>
    <author initials="T." surname="Cholez" fullname="Thibault Cholez">
      <organization></organization>
    </author>
    <author initials="O." surname="Bettan" fullname="Olivier Bettan">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/TNSM.2019.2933155"/>
<annotation>IEEE Transactions on Network and Service Management, 16(3), pp. 842-856</annotation></reference>
<reference anchor="Hoo18" >
  <front>
    <title>Updated Taxonomy for the Network and Service Management Research Field</title>
    <author initials="J. van der" surname="Hooft" fullname="Jeroen van der Hooft">
      <organization></organization>
    </author>
    <author initials="M." surname="Claeys" fullname="Maxim Claeys">
      <organization></organization>
    </author>
    <author initials="N." surname="Bouten" fullname="Niels Bouten">
      <organization></organization>
    </author>
    <author initials="T." surname="Wauters" fullname="Tim Wauters">
      <organization></organization>
    </author>
    <author initials="J." surname="Schönwälder" fullname="Jürgen Schönwälder">
      <organization></organization>
    </author>
    <author initials="A." surname="Pras" fullname="Aiko Pras">
      <organization></organization>
    </author>
    <author initials="B." surname="Stiller" fullname="Burkhard Stiller">
      <organization></organization>
    </author>
    <author initials="M." surname="Charalambides" fullname="Marinos Charalambides">
      <organization></organization>
    </author>
    <author initials="R." surname="Badonnel" fullname="Rémi Badonnel">
      <organization></organization>
    </author>
    <author initials="J." surname="Serrat" fullname="Joan Serrat">
      <organization></organization>
    </author>
    <author initials="C. R. P. dos" surname="Santos" fullname="Carlos Raniery Paula dos Santos">
      <organization></organization>
    </author>
    <author initials="F. D." surname="Turck" fullname="Filip De Turck">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1007/s10922-017-9443-y"/>
<annotation>Journal of Network and Systems Management (JNSM), 26(3), pp. 790-808</annotation></reference>
<reference anchor="Bou18" >
  <front>
    <title>A comprehensive survey on machine learning for networking: evolution, applications and research opportunities</title>
    <author initials="R." surname="Boutaba" fullname="Raouf Boutaba">
      <organization></organization>
    </author>
    <author initials="M. A." surname="Salahuddin" fullname="Mohammad A. Salahuddin">
      <organization></organization>
    </author>
    <author initials="N." surname="Limam" fullname="Noura Limam">
      <organization></organization>
    </author>
    <author initials="S." surname="Ayoubi" fullname="Sara Ayoubi">
      <organization></organization>
    </author>
    <author initials="N." surname="Shahriar" fullname="Nashid Shahriar">
      <organization></organization>
    </author>
    <author initials="F." surname="Estrada-Solano" fullname="Felipe Estrada-Solano">
      <organization></organization>
    </author>
    <author initials="O. M." surname="Caicedo" fullname="Oscar M. Caicedo">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1186/s13174-018-0087-2"/>
<annotation>Journal of Internet Services and Applications, 9, article 16</annotation></reference>
<reference anchor="Kaf19" >
  <front>
    <title>Automation of 5G Network Slice Control Functions with Machine Learning</title>
    <author initials="V. P." surname="Kafle" fullname="V. P. Kafle">
      <organization></organization>
    </author>
    <author initials="P." surname="Martinez-Julia" fullname="P. Martinez-Julia">
      <organization></organization>
    </author>
    <author initials="T." surname="Miyazawa" fullname="T. Miyazawa">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/MCOMSTD.001.1900010"/>
<annotation>IEEE Communications Standards Magazine, 3(3), pp. 54-62</annotation></reference>
<reference anchor="Yan20" >
  <front>
    <title>Artificial-Intelligence-Enabled Intelligent 6G Networks</title>
    <author initials="H." surname="Yang" fullname="H. Yang">
      <organization></organization>
    </author>
    <author initials="A." surname="Alphones" fullname="A. Alphones">
      <organization></organization>
    </author>
    <author initials="Z." surname="Xiong" fullname="Z. Xiong">
      <organization></organization>
    </author>
    <author initials="D." surname="Niyato" fullname="D. Niyato">
      <organization></organization>
    </author>
    <author initials="J." surname="Zhao" fullname="J. Zhao">
      <organization></organization>
    </author>
    <author initials="K." surname="Wu" fullname="K. Wu">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/MNET.011.2000195"/>
<annotation>IEEE Network, 34(6), pp. 272-280</annotation></reference>
<reference anchor="Lop20" >
  <front>
    <title>Priority Flow Admission and Routing in SDN: Exact and Heuristic Approaches</title>
    <author initials="J." surname="López" fullname="J. López">
      <organization></organization>
    </author>
    <author initials="M." surname="Labonne" fullname="M. Labonne">
      <organization></organization>
    </author>
    <author initials="C." surname="Poletti" fullname="C. Poletti">
      <organization></organization>
    </author>
    <author initials="D." surname="Belabed" fullname="D. Belabed">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/NCA51143.2020.9306725"/>
<annotation>IEEE International Symposium on Network Computing and Applications (NCA)</annotation></reference>
<reference anchor="Sen04" >
  <front>
    <title>Accurate, scalable in-network identification of p2p traffic using application signatures</title>
    <author initials="S." surname="Sen" fullname="Subhabrata Sen">
      <organization></organization>
    </author>
    <author initials="O." surname="Spatscheck" fullname="Oliver Spatscheck">
      <organization></organization>
    </author>
    <author initials="D." surname="Wang" fullname="Dongmei Wang">
      <organization></organization>
    </author>
    <date year="2004"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/988672.988742"/>
<annotation>ACM International Conference on World Wide Web (WWW)</annotation></reference>
<reference anchor="Rin17" >
  <front>
    <title>IP2Vec: Learning Similarities Between IP Addresses</title>
    <author initials="M." surname="Ring" fullname="M. Ring">
      <organization></organization>
    </author>
    <author initials="A." surname="Dallmann" fullname="A. Dallmann">
      <organization></organization>
    </author>
    <author initials="D." surname="Landes" fullname="D. Landes">
      <organization></organization>
    </author>
    <author initials="A." surname="Hotho" fullname="A. Hotho">
      <organization></organization>
    </author>
    <date year="2017"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/ICDMW.2017.93"/>
<annotation>IEEE International Conference on Data Mining Workshops (ICDMW)</annotation></reference>
<reference anchor="Evr19" >
  <front>
    <title>port2dist: Semantic Port Distances for Network Analytics</title>
    <author initials="L." surname="Evrard" fullname="Laurent Evrard">
      <organization></organization>
    </author>
    <author initials="J." surname="François" fullname="Jérôme François">
      <organization></organization>
    </author>
    <author initials="J.-N." surname="Colin" fullname="Jean-Noël Colin">
      <organization></organization>
    </author>
    <author initials="F." surname="Beck" fullname="Frédéric Beck">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
<annotation>IFIP/IEEE Symposium on Integrated Network and Service Management (IM)</annotation></reference>
<reference anchor="Sol20" >
  <front>
    <title>A Graph Neural Network Approach for Scalable and Dynamic IP Similarity in Enterprise Networks</title>
    <author initials="H. M." surname="Soliman" fullname="Hazem M. Soliman">
      <organization></organization>
    </author>
    <author initials="G." surname="Salmon" fullname="Geoff Salmon">
      <organization></organization>
    </author>
    <author initials="D." surname="Sovilj" fullname="Dusan Sovilj">
      <organization></organization>
    </author>
    <author initials="M." surname="Rao" fullname="Mohan Rao">
      <organization></organization>
    </author>
    <date year="2020"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/CloudNet51028.2020.9335789"/>
<annotation>IEEE International Conference on Cloud Networking (CloudNet)</annotation></reference>
<reference anchor="Sco11" >
  <front>
    <title>On Measuring the Similarity of Network Hosts: Pitfalls, New Metrics, and Empirical Analyses</title>
    <author initials="S. E." surname="Coull" fullname="Scott E. Coull">
      <organization></organization>
    </author>
    <author initials="F." surname="Monrose" fullname="Fabian Monrose">
      <organization></organization>
    </author>
    <author initials="M." surname="Bailey" fullname="Michael Bailey">
      <organization></organization>
    </author>
    <date year="2011"/>
  </front>
<annotation>Network and Distributed System Security Symposium (NDSS)</annotation></reference>
<reference anchor="Xie19" >
  <front>
    <title>A survey of machine learning techniques applied to software defined networking (SDN): Research issues and challenges</title>
    <author initials="J." surname="Xie" fullname="Junfeng Xie">
      <organization></organization>
    </author>
    <author initials="F. R." surname="Yu" fullname="F. Richard Yu">
      <organization></organization>
    </author>
    <author initials="T." surname="Huang" fullname="Tao Huang">
      <organization></organization>
    </author>
    <author initials="R." surname="Xie" fullname="Renchao Xie">
      <organization></organization>
    </author>
    <author initials="J." surname="Liu" fullname="Jiang Liu">
      <organization></organization>
    </author>
    <author initials="C." surname="Wang" fullname="Chenmeng Wang">
      <organization></organization>
    </author>
    <author initials="Y." surname="Liu" fullname="Yunjie Liu">
      <organization></organization>
    </author>
    <date year="2019"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/COMST.2018.2866942"/>
<annotation>IEEE Communications Surveys &amp; Tutorials, 21(1), pp. 393-430</annotation></reference>
<reference anchor="Yu14" >
  <front>
    <title>Distributed and collaborative traffic monitoring in software defined networks</title>
    <author initials="Y." surname="Yu" fullname="Ye Yu">
      <organization></organization>
    </author>
    <author initials="C." surname="Qian" fullname="Chen Qian">
      <organization></organization>
    </author>
    <author initials="X." surname="Li" fullname="Xin Li">
      <organization></organization>
    </author>
    <date year="2014"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/2620728.2620739"/>
<annotation>ACM Workshop on Hot Topics in Software Defined Networking (HotSDN)</annotation></reference>
<reference anchor="Dij18" >
  <front>
    <title>A Survey of Network Traffic Anonymisation Techniques and Implementations</title>
    <author initials="N. V." surname="Dijkhuizen" fullname="Niels Van Dijkhuizen">
      <organization></organization>
    </author>
    <author initials="J. V. D." surname="Ham" fullname="Jeroen Van Der Ham">
      <organization></organization>
    </author>
    <date year="2018"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3182660"/>
<annotation>ACM Computing Surveys, 51(3), article 52</annotation></reference>
<reference anchor="Sch21" >
  <front>
    <title>Distributed Online Service Coordination Using Deep Reinforcement Learning</title>
    <author initials="S." surname="Schneider" fullname="Stefan Schneider">
      <organization></organization>
    </author>
    <author initials="H." surname="Qarawlus" fullname="Haydar Qarawlus">
      <organization></organization>
    </author>
    <author initials="H." surname="Karl" fullname="Holger Karl">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
  <seriesInfo name="DOI" value="10.1109/ICDCS51616.2021.00058"/>
<annotation>IEEE International Conference on Distributed Computing Systems (ICDCS)</annotation></reference>
<reference anchor="Hui22" >
  <front>
    <title>Knowledge Enhanced GAN for IoT Traffic Generation</title>
    <author initials="S." surname="Hui" fullname="Shuodi Hui">
      <organization></organization>
    </author>
    <author initials="H." surname="Wang" fullname="Huandong Wang">
      <organization></organization>
    </author>
    <author initials="Z." surname="Wang" fullname="Zhenhua Wang">
      <organization></organization>
    </author>
    <author initials="X." surname="Yang" fullname="Xinghao Yang">
      <organization></organization>
    </author>
    <author initials="Z." surname="Liu" fullname="Zhongjin Liu">
      <organization></organization>
    </author>
    <author initials="D." surname="Jin" fullname="Depeng Jin">
      <organization></organization>
    </author>
    <author initials="Y." surname="Li" fullname="Yong Li">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/3485447.3511976"/>
<annotation>ACM Web Conference 2022 (WWW)</annotation></reference>
<reference anchor="Sil16" >
  <front>
    <title>Mastering the game of Go with deep neural networks and tree search</title>
    <author initials="D." surname="Silver" fullname="David Silver">
      <organization></organization>
    </author>
    <author initials="A." surname="Huang" fullname="Aja Huang">
      <organization></organization>
    </author>
    <author initials="C. J." surname="Maddison" fullname="Christopher J. Maddison">
      <organization></organization>
    </author>
    <author initials="A." surname="Guez" fullname="Arthur Guez">
      <organization></organization>
    </author>
    <author initials="L." surname="Sifre" fullname="Laurent Sifre">
      <organization></organization>
    </author>
    <author initials="G. van den" surname="Driessche" fullname="George van den Driessche">
      <organization></organization>
    </author>
    <author initials="J." surname="Schrittwieser" fullname="Julian Schrittwieser">
      <organization></organization>
    </author>
    <author initials="I." surname="Antonoglou" fullname="Ioannis Antonoglou">
      <organization></organization>
    </author>
    <author initials="V." surname="Panneershelvam" fullname="Veda Panneershelvam">
      <organization></organization>
    </author>
    <author initials="M." surname="Lanctot" fullname="Marc Lanctot">
      <organization></organization>
    </author>
    <author initials="S." surname="Dieleman" fullname="Sander Dieleman">
      <organization></organization>
    </author>
    <author initials="D." surname="Grewe" fullname="Dominik Grewe">
      <organization></organization>
    </author>
    <author initials="J." surname="Nham" fullname="John Nham">
      <organization></organization>
    </author>
    <author initials="N." surname="Kalchbrenner" fullname="Nal Kalchbrenner">
      <organization></organization>
    </author>
    <author initials="I." surname="Sutskever" fullname="Ilya Sutskever">
      <organization></organization>
    </author>
    <author initials="T." surname="Lillicrap" fullname="Timothy Lillicrap">
      <organization></organization>
    </author>
    <author initials="M." surname="Leach" fullname="Madeleine Leach">
      <organization></organization>
    </author>
    <author initials="K." surname="Kavukcuoglu" fullname="Koray Kavukcuoglu">
      <organization></organization>
    </author>
    <author initials="T." surname="Graepel" fullname="Thore Graepel">
      <organization></organization>
    </author>
    <author initials="D." surname="Hassabis" fullname="Demis Hassabis">
      <organization></organization>
    </author>
    <date year="2016"/>
  </front>
  <seriesInfo name="DOI" value="10.1038/nature16961"/>
<annotation>Nature, 529(7587), pp. 484-489</annotation></reference>
<reference anchor="Nas21" >
  <front>
    <title>Defeating DNN-Based Traffic Analysis Systems in Real-Time With Blind Adversarial Perturbations</title>
    <author initials="M." surname="Nasr" fullname="M. Nasr">
      <organization></organization>
    </author>
    <author initials="A." surname="Bahramali" fullname="A. Bahramali">
      <organization></organization>
    </author>
    <author initials="A." surname="Houmansadr" fullname="A. Houmansadr">
      <organization></organization>
    </author>
    <date year="2021"/>
  </front>
<annotation>USENIX Security Symposium</annotation></reference>
<reference anchor="Jmi22" >
  <front>
    <title>Adversarial machine learning for network intrusion detection: A comparative study</title>
    <author initials="H." surname="Jmila" fullname="H. Jmila">
      <organization></organization>
    </author>
    <author initials="M. I." surname="Khedher" fullname="M. I. Khedher">
      <organization></organization>
    </author>
    <date year="2022"/>
  </front>
  <seriesInfo name="DOI" value="10.1016/j.comnet.2022.109073"/>
<annotation>Computer Networks, Elsevier, vol. 214</annotation></reference>
<reference anchor="Fre15" >
  <front>
    <title>Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures</title>
    <author initials="M." surname="Fredrikson" fullname="M. Fredrikson">
      <organization></organization>
    </author>
    <author initials="S." surname="Jha" fullname="S. Jha">
      <organization></organization>
    </author>
    <author initials="T." surname="Ristenpart" fullname="T. Ristenpart">
      <organization></organization>
    </author>
    <date year="2015"/>
  </front>
  <seriesInfo name="DOI" value="10.1145/2810103.2813677"/>
<annotation>ACM SIGSAC Conference on Computer and Communications Security (CCS)</annotation></reference>
<reference anchor="Bis23" >
  <front>
    <title>Hyperparameter optimization: Foundations, algorithms, best practices, and open challenges</title>
    <author initials="B." surname="Bischl" fullname="B. Bischl">
      <organization></organization>
    </author>
    <author initials="M." surname="Binder" fullname="M. Binder">
      <organization></organization>
    </author>
    <author initials="M." surname="Lang" fullname="M. Lang">
      <organization></organization>
    </author>
    <author initials="T." surname="Pielok" fullname="T. Pielok">
      <organization></organization>
    </author>
    <author initials="J." surname="Richter" fullname="J. Richter">
      <organization></organization>
    </author>
    <author initials="S." surname="Coors" fullname="S. Coors">
      <organization></organization>
    </author>
    <author initials="J." surname="Thomas" fullname="J. Thomas">
      <organization></organization>
    </author>
    <author initials="T." surname="Ullmann" fullname="T. Ullmann">
      <organization></organization>
    </author>
    <author initials="M." surname="Becker" fullname="M. Becker">
      <organization></organization>
    </author>
    <author initials="A.-L." surname="Boulesteix" fullname="A.-L. Boulesteix">
      <organization></organization>
    </author>
    <author initials="D." surname="Deng" fullname="D. Deng">
      <organization></organization>
    </author>
    <author initials="M." surname="Lindauer" fullname="M. Lindauer">
      <organization></organization>
    </author>
    <date year="2023"/>
  </front>
  <seriesInfo name="DOI" value="10.1002/widm.1484"/>
<annotation>Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 13(2)</annotation></reference>


    </references>


<section numbered="false" anchor="acknowledgments"><name>Acknowledgments</name>

<t>This document is the result of a collective work. Authors of this document are the main contributors and the editors, but contributions have been also received from the following people we acknowledge Laurent Ciavaglia, Felipe Alencar Lopes, Abdelkader Lahamdi, Albert Cabellos, José Suárez-Varela, Marinos Charalambides, Ramin Sadre, Pedro Martinez-Julia and Flavio Esposito.</t>

<t>This document was also partially supported by project AI@EDGE, funded from the European Union's Horizon 2020 H2020-ICT-52 call for projects, under grant agreement no. 101015922. This research is funded in part, by the Luxembourg National Research Fund (FNR), grant reference C23/IS/18088425/COCTEL. 
The views expressed in this document do not necessarily reflect those of Canada Post.</t>

</section>


  </back>

<!-- ##markdown-source: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-->

</rfc>

