Skip to main content

Research Challenges in Coupling Artificial Intelligence and Network Management
draft-irtf-nmrg-ai-challenges-06

Document Type Active Internet-Draft (nmrg RG)
Authors Jérôme François , Alexander Clemm , Dimitri Papadimitriou , Stenio Fernandes , Stefan Schneider
Last updated 2026-09-08 (Latest revision 2026-07-06)
Replaces draft-francois-nmrg-ai-challenges
RFC stream Internet Research Task Force (IRTF)
Intended RFC status Informational
Formats
IETF conflict review conflict-review-irtf-nmrg-ai-challenges
Additional resources Mailing list discussion
Stream IRTF state In IESG Review
Consensus boilerplate Yes
Document shepherd Jéferson Campos Nobre
Shepherd write-up Show Last changed 2026-03-30
IESG IESG state I-D Exists
Telechat date (None)
Responsible AD (None)
Send notices to Laurent.Ciavaglia@nokia.com, jcnobre@inf.ufrgs.br
draft-irtf-nmrg-ai-challenges-06
Internet Research Task Force                                 J. François
Internet-Draft                        University of Luxembourg and Inria
Intended status: Informational                                  A. Clemm
Expires: 7 January 2027                                      Independent
                                                        D. Papadimitriou
                                           3NLab Belgium Research Center
                                                            S. Fernandes
                                                             Canada Post
                                                            S. Schneider
                                  Digital Railway (DSD) at Deutsche Bahn
                                                             6 July 2026

  Research Challenges in Coupling Artificial Intelligence and Network
                               Management
                    draft-irtf-nmrg-ai-challenges-06

Abstract

   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.

   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.

   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.

François, et al.         Expires 7 January 2027                 [Page 1]
Internet-Draft     Coupling AI and network management          July 2026

Status of This Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at https://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on 7 January 2027.

Copyright Notice

   Copyright (c) 2026 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents (https://trustee.ietf.org/
   license-info) in effect on the date of publication of this document.
   Please review these documents carefully, as they describe your rights
   and restrictions with respect to this document.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   3
   2.  Acronyms  . . . . . . . . . . . . . . . . . . . . . . . . . .   5
   3.  Difficult problems in network management  . . . . . . . . . .   6
   4.  High-level challenges in adopting AI in NM  . . . . . . . . .  10
   5.  AI techniques and network management  . . . . . . . . . . . .  13
     5.1.  Problem type and mapping  . . . . . . . . . . . . . . . .  13
       5.1.1.  Sub-challenge: Suitable Approach for Given Input  . .  13
       5.1.2.  Sub-challenge: Suitable Approach for Desired
               Output  . . . . . . . . . . . . . . . . . . . . . . .  14
       5.1.3.  Sub-challenge: Tailoring the AI Approach to the Given
               Problem . . . . . . . . . . . . . . . . . . . . . . .  15
     5.2.  Performance of produced models  . . . . . . . . . . . . .  16
     5.3.  Lightweight AI  . . . . . . . . . . . . . . . . . . . . .  18
     5.4.  Distributed AI  . . . . . . . . . . . . . . . . . . . . .  20
       5.4.1.  Network management for efficient distributed AI . . .  20
       5.4.2.  Distributed AI for network management . . . . . . . .  21
     5.5.  AI for planning of actions  . . . . . . . . . . . . . . .  21
   6.  Network data as input for ML algorithms . . . . . . . . . . .  23

François, et al.         Expires 7 January 2027                 [Page 2]
Internet-Draft     Coupling AI and network management          July 2026

     6.1.  Data for AI-based NM solutions  . . . . . . . . . . . . .  24
     6.2.  Data collection . . . . . . . . . . . . . . . . . . . . .  25
     6.3.  Usable data . . . . . . . . . . . . . . . . . . . . . . .  26
   7.  Acceptability of AI . . . . . . . . . . . . . . . . . . . . .  28
     7.1.  Explainability of Network-AI products . . . . . . . . . .  28
     7.2.  AI-based products and algorithms in production systems  .  29
     7.3.  AI with humans in the loop  . . . . . . . . . . . . . . .  31
   8.  Conclusions and Future Research Directions  . . . . . . . . .  32
   9.  Security considerations . . . . . . . . . . . . . . . . . . .  33
     9.1.  AI-based security solutions . . . . . . . . . . . . . . .  33
     9.2.  Security of AI  . . . . . . . . . . . . . . . . . . . . .  34
     9.3.  Relevance of AI-based outputs . . . . . . . . . . . . . .  35
   10. IANA Considerations . . . . . . . . . . . . . . . . . . . . .  35
   11. References  . . . . . . . . . . . . . . . . . . . . . . . . .  35
     11.1.  Normative References . . . . . . . . . . . . . . . . . .  35
     11.2.  Informative References . . . . . . . . . . . . . . . . .  36
   Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . .  48
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  48

1.  Introduction

   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 [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.

   Research in NM and networking has been very active in applied ML
   [Bou18] and has enabled progress in areas such as network security
   [Ke23], Software-Defined Networks (SDN) [Ham21], and vehicular
   networks [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 [Sha25].

   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

François, et al.         Expires 7 January 2027                 [Page 3]
Internet-Draft     Coupling AI and network management          July 2026

   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.

   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.

   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 [Kaf19].

   However, not all computational problems in network management are NP-
   hard.  Due to real-time constraints, some involve very short control
   loops that require both rapid decisions and the ability to rapidly
   adapt to new situations and different contexts.  So, even in that
   case, time is critical, and approximate solutions are usually more
   acceptable.  Again, it is where AI can be beneficial.  Actually,
   expert systems are AI systems [Ste92] but this kind of rule-based
   systems are not designed to scale with the volume and heterogeneity
   of data we can collect in a network today.  In contrast, ML is more
   efficient to automatically learn abstract representations of the
   rules, which can be eventually updated.

François, et al.         Expires 7 January 2027                 [Page 4]
Internet-Draft     Coupling AI and network management          July 2026

   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
   [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.

   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.

   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.

   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 [czb20]
   [Yan20], as they require unprecedented levels of performance in terms
   of throughput, latency, mobility, security, and more.

2.  Acronyms

   *  AI: Artificial Intelligence

François, et al.         Expires 7 January 2027                 [Page 5]
Internet-Draft     Coupling AI and network management          July 2026

   *  CNN: Convolutional Neural Network

   *  FL: Federated Learning

   *  GAN: Generative Adversarial Network

   *  GNN: Graph Neural Network

   *  IBN: Intent-Based Networking

   *  LLM: Large Language Model

   *  LSTM: Long Short-Term Memory

   *  MAE: Mean Absolute Error

   *  ML: Machine Learning

   *  MILP: Mixed-Integer Linear Programming

   *  MLP: Multilayer Perceptron

   *  MSE: Mean Squared Error

   *  NM: Network Management

   *  RL: Reinforcement Learning

   *  SDN: Software-Defined Networks

   *  SFC: Service Function Chaining

   *  SVM: Support Vector Machine

   *  VNF: Virtual Network Function

3.  Difficult problems in network management

   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:

François, et al.         Expires 7 January 2027                 [Page 6]
Internet-Draft     Coupling AI and network management          July 2026

   *  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).

   *  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.

   *  C3: Real-Time Constraints -- 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.

   *  C4: Data Dependency -- Processing and analyzing massive volumes of
      network telemetry to derive accurate management decisions, which
      requires dealing with compound data-plane challenges ranging from
      data heterogeneity and incomplete or missing datasets to high data
      velocity.

   *  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.

   *  C6: Economics -- 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.

François, et al.         Expires 7 January 2027                 [Page 7]
Internet-Draft     Coupling AI and network management          July 2026

     +==================+============================================+
     | Characteristic   | Relevant Document Sections / Gaps          |
     | Tag & Short Name |                                            |
     +==================+============================================+
     | C1: Fundamental  | Section 5.1 (Problem Mapping); Section 5.5 |
     | Algorithmic      | (Action Planning); Section 5.4             |
     | Complexity       | (Distributed AI)                           |
     +------------------+--------------------------------------------+
     | C2: Uncertainty  | Section 5.5 (Action Planning); Section 5.2 |
     | and              | (Performance of produced models);          |
     | Unpredictability | Section 9.2 (Security of AI); Section 9.3  |
     |                  | (Relevance of AI-based outputs)            |
     +------------------+--------------------------------------------+
     | C3: Real-Time    | Section 5.3 (Lightweight AI); Section 7.2  |
     | Constraints and  | (Production Systems); Section 5.2          |
     | Determinism      | (Performance of produced models)           |
     +------------------+--------------------------------------------+
     | C4: Data         | Section 6.1 (Network data as input for ML  |
     | Dependency and   | algorithms); Section 6.2 (Data             |
     | Heterogeneity    | collection); Section 6.3 (Usable data)     |
     +------------------+--------------------------------------------+
     | C5: Integration  | Section 7.3 (AI with humans in the loop);  |
     | with Existing    | Section 5.4.1 (NM for AI); Section 7.1     |
     | Practice         | (Explainability)                           |
     +------------------+--------------------------------------------+
     | C6: Economics    | Section 5.3 (Lightweight AI); Section 7.2  |
     | and              | (Production Systems)                       |
     | Computational    |                                            |
     | Footprint        |                                            |
     +------------------+--------------------------------------------+

                      Table 1: Problem characteristics

   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:

   *  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,

François, et al.         Expires 7 January 2027                 [Page 8]
Internet-Draft     Coupling AI and network management          July 2026

      operations research techniques have been extensively used in
      literature along with AI methods [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 [Sin22].  As such, this problem can be considered as close
      to big data problems with some of the different Vs: volume,
      velocity, variety, value…

   *  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 [Sen04].  To handle applications now encapsulated in
      encrypted web or VPN traffic, various ML-based approaches have
      been thus adopted [Bri19][Naj24][Wan24][Ake24] including LLMs
      [Gin24].

   *  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[Mut24].

   *  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 [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) [Ami24].

François, et al.         Expires 7 January 2027                 [Page 9]
Internet-Draft     Coupling AI and network management          July 2026

   *  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 [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 [Col22].

   *  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 [Att23].

   *  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 [Vin21][Sul23].

4.  High-level challenges in adopting AI in NM

   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 extansive research in this direction but has
   not widely adopted those solutions to operate real networks.  There
   are several obstacles.

François, et al.         Expires 7 January 2027                [Page 10]
Internet-Draft     Coupling AI and network management          July 2026

   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 [Sha25].

   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 Section 5.

   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
   [Sco11].  However, more advanced techniques can be proposed to embed
   representation of network entities rather than pure encoding as
   illustrated in [Rin17][Evr19][Sol20].

   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 [Liu22].  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 [Hoa21].  Intelligent
   data analysis and the need to maintain privacy are in many ways

François, et al.         Expires 7 January 2027                [Page 11]
Internet-Draft     Coupling AI and network management          July 2026

   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 [Liu22b].  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 Section 6.

   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 [Liu21] and
   synchronization to maintain a coherence over a distributed system.

   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 [Sen24].

   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 Section 7.

François, et al.         Expires 7 January 2027                [Page 12]
Internet-Draft     Coupling AI and network management          July 2026

5.  AI techniques and network management

5.1.  Problem type and mapping

   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 [Mus18], [Xie18].
   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).

   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.

5.1.1.  Sub-challenge: Suitable Approach for Given Input

   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.

   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 [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

François, et al.         Expires 7 January 2027                [Page 13]
Internet-Draft     Coupling AI and network management          July 2026

   (transformers) are great for sequential data like time series or
   text.  This evolution leads to the era of LLMs also impacting
   research in networking [Hua25][Wu24].

   It is worth noting that Graph Neural Networks (GNNs) can incorporate
   and consider the graph-structured input, which is very useful in
   network management [Jie22], e.g., to represent the network topology.

   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
   [Ham23], e.g., with both CNNs and LSTMs, and multiple separate AI
   approaches can be used as an ensemble to combine their strengths
   [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.

5.1.2.  Sub-challenge: Suitable Approach for Desired Output

   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.

   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
   [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 [Liu22b] 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

François, et al.         Expires 7 January 2027                [Page 14]
Internet-Draft     Coupling AI and network management          July 2026

   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 [Urb23].

   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 [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.

   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.

5.1.3.  Sub-challenge: Tailoring the AI Approach to the Given Problem

   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.

   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

François, et al.         Expires 7 January 2027                [Page 15]
Internet-Draft     Coupling AI and network management          July 2026

   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.

   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 [Arz21].

5.2.  Performance of produced models

   From a general point of view, any AI technique will produce results
   with a certain level of quality.  This leads to two 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?

   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).

François, et al.         Expires 7 January 2027                [Page 16]
Internet-Draft     Coupling AI and network management          July 2026

   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.

   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.

   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 [Beg19].

François, et al.         Expires 7 January 2027                [Page 17]
Internet-Draft     Coupling AI and network management          July 2026

   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 [Hua19].

   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).

   Finally, question (4) refers to the ability of an AI solution to
   remain efficient and eventually improve over time [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 [Sha21b],
   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.

5.3.  Lightweight AI

   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.

François, et al.         Expires 7 January 2027                [Page 18]
Internet-Draft     Coupling AI and network management          July 2026

   Recent advances in network programmability facilitate the programming
   of specific tasks at the data-plane level.  P4 [Bos14] is widely used
   today for many tasks including firewalling [Dat18] or bandwidth
   management [Che19].  P4 is generally agnostic to specific hardware.
   It is based on the RMT (Reconfigurable Match Table) architectural
   model [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.

   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) [Mcc93] programs provide greater control over packet
   processing in OVS [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.

   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 [Liu16][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 [Hau23] with a few attempts to propose generic programs to
   be reusable or composed as a kind of libraries such as [Zha24]
   leveraging quantization, [Jos21][Jos22] relying on pre-computed
   lookup tables of real-value functions or [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 [Gup18].

François, et al.         Expires 7 January 2027                [Page 19]
Internet-Draft     Coupling AI and network management          July 2026

5.4.  Distributed AI

   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.

5.4.1.  Network management for efficient distributed AI

   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 [Lin21].  For example, in the context of optimizing FL
   [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 [Raj24].  It paves the way to
   research on methods to support AI-aware NM or network-aware AI
   applications or a combination of both.

François, et al.         Expires 7 January 2027                [Page 20]
Internet-Draft     Coupling AI and network management          July 2026

5.4.2.  Distributed AI for network management

   For network management applications relying on distributed AI,
   challenges from Section 5.4.1 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 [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 Section 5.4.1, since the network is both part of
   the AI pipeline resources and the managed object through AI.

   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.

5.5.  AI for planning of actions

   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 Section 3.

François, et al.         Expires 7 January 2027                [Page 21]
Internet-Draft     Coupling AI and network management          July 2026

   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 [Ahm21].

   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 [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.

   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 [Abd24].

   A much-publicized leap in AI has been the development of AlphaGo
   [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.

   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.

François, et al.         Expires 7 January 2027                [Page 22]
Internet-Draft     Coupling AI and network management          July 2026

   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 [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 [Cor22].

   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 [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 [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
   [Sha24][Ans22].  Generally, it is also an open problem to make the
   use of AI more acceptable, as highlighted in the dedicated section.

6.  Network data as input for ML algorithms

   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.

   Thus, the challenge of data management is multifaceted, as detailed
   in the next subsections.

François, et al.         Expires 7 January 2027                [Page 23]
Internet-Draft     Coupling AI and network management          July 2026

6.1.  Data for AI-based NM solutions

   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 [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 [Jie22].
   As an example, graph-based representations are considered practically
   efficient due to their intrinsic structural similarities with network
   data, as illustrated in [Bar23] for attack detection.

   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 [Mek24][Fua24][Dze23].
   Secondly, similar to the problem of mapping AI algorithms with NM
   problems in section Section 5.1, 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 [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.

   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

François, et al.         Expires 7 January 2027                [Page 24]
Internet-Draft     Coupling AI and network management          July 2026

   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 [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.

6.2.  Data collection

   Once defined, the second problem to address is the collection of
   data.  Monitoring frameworks have been developed for many years such
   as IPFIX [RFC7011] and more recently with SDN-based monitoring
   solutions [Yu14][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 [Bri19] or by involving end-users with
   crowd-sourcing [Hir15] and distributed monitoring.  Also, the data-
   mesh concept [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).

   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

François, et al.         Expires 7 January 2027                [Page 25]
Internet-Draft     Coupling AI and network management          July 2026

   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 [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.

   In parallel to the continuously increasing dynamic behavior of
   networks and complexity of traffic, there is a trend towards more
   user traffic processing customization [RFC8986][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 [Tan20b].  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.

6.3.  Usable data

   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 [Dij19] 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/

François, et al.         Expires 7 January 2027                [Page 26]
Internet-Draft     Coupling AI and network management          July 2026

   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 [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.

   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 [Dub24].

   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,

François, et al.         Expires 7 January 2027                [Page 27]
Internet-Draft     Coupling AI and network management          July 2026

   intrusion detection systems, etc.  The usage of AI, Generative
   Adversarial Networks (GANs), in this area [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 [Jia24] or using a real
   testbed guided by AI to create challenging scenarios for validation
   purposes [Ans25].

7.  Acceptability of AI

   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.

7.1.  Explainability of Network-AI products

   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.

   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.

François, et al.         Expires 7 January 2027                [Page 28]
Internet-Draft     Coupling AI and network management          July 2026

   The networking field has already produced a set of well-established
   network management algorithms and methods, with clear performance
   guarantees and troubleshooting mechanisms [Rex06][Kr14].  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 [XAI].

   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.

   There exist several ML-based methods that are human-understandable,
   although not widely used today.  For instance, [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 [Zil20][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.)

7.2.  AI-based products and algorithms in production systems

   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?

   Traditionally, AI-based models have been trained in the same scenario
   where they operate[Val17][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

François, et al.         Expires 7 January 2027                [Page 29]
Internet-Draft     Coupling AI and network management          July 2026

   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.

   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.

   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
   [gnn1][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 [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.

François, et al.         Expires 7 January 2027                [Page 30]
Internet-Draft     Coupling AI and network management          July 2026

   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.

   This shares similarities with problems described in Section 5.4 and
   particularly emphasizes the need for network environments to provide
   interfaces and descriptions suitable for AI solutions to be properly
   instantiated and configured.

   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) [Mor18].

7.3.  AI with humans in the loop

   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 [Wu22], which could be useful for
   different aspects of network management.

François, et al.         Expires 7 January 2027                [Page 31]
Internet-Draft     Coupling AI and network management          July 2026

   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 SageMake).  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.

   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
   [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.

8.  Conclusions and Future Research Directions

   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.

   Key takeaways include:

   *  NM problems often require network-specific solutions rather than
      direct imports from other AI domains.

   *  Data quality, representation, and privacy-aware sharing are
      fundamental enablers for effective AI-based NM.

   *  Lightweight and distributed AI architectures are needed to meet
      timing, resource, and deployment constraints.

François, et al.         Expires 7 January 2027                [Page 32]
Internet-Draft     Coupling AI and network management          July 2026

   *  Seamless integration with existing NM processes, explainability,
      and human-in-the-loop interaction are essential for acceptability.

   *  Practical deployment also requires evaluation methods, safeguards,
      and mechanisms for models to generalize and adapt over time.

9.  Security considerations

   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.

9.1.  AI-based security solutions

   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 Section 7.1.

   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

François, et al.         Expires 7 January 2027                [Page 33]
Internet-Draft     Coupling AI and network management          July 2026

   Hyper-Parameter Optimization (HPO) [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 Section 5.1.3 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.

   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 Section 6 of this document fully addresses the concerns
   related to data in regard to NM problems.

9.2.  Security of AI

   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.

   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 [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 [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
   [Fre15] raises concern regarding privacy.

François, et al.         Expires 7 January 2027                [Page 34]
Internet-Draft     Coupling AI and network management          July 2026

   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.

9.3.  Relevance of AI-based outputs

   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.

10.  IANA Considerations

   This document has no IANA actions.

11.  References

11.1.  Normative References

   [RFC7011]  Claise, B., Ed., Trammell, B., Ed., and P. Aitken,
              "Specification of the IP Flow Information Export (IPFIX)
              Protocol for the Exchange of Flow Information", STD 77,
              RFC 7011, DOI 10.17487/RFC7011, September 2013,
              <https://www.rfc-editor.org/info/rfc7011>.

   [RFC8986]  Filsfils, C., Ed., Camarillo, P., Ed., Leddy, J., Voyer,
              D., Matsushima, S., and Z. Li, "Segment Routing over IPv6
              (SRv6) Network Programming", RFC 8986,
              DOI 10.17487/RFC8986, February 2021,
              <https://www.rfc-editor.org/info/rfc8986>.

François, et al.         Expires 7 January 2027                [Page 35]
Internet-Draft     Coupling AI and network management          July 2026

   [RFC9315]  Clemm, A., Ciavaglia, L., Granville, L. Z., and J.
              Tantsura, "Intent-Based Networking - Concepts and
              Definitions", RFC 9315, DOI 10.17487/RFC9315, October
              2022, <https://www.rfc-editor.org/info/rfc9315>.

11.2.  Informative References

   [Abd10]    Jalil, K. A., Kamarudin, M. H., and M. N. Masrek, "A
              Diagnosis Expert System for Network Traffic Management",
              2010.  IEEE international conference on networking and
              information technology

   [Abd24]    Bouroudi, A., Outtagarts, A., and Y. Hadjadj-Aoul, "A
              dynamic AI-based algorithm selection for Virtual Network
              Embedding", 2024.  Springer Annals of Telecommunication

   [Ahm21]    Ahmad, S. and A. H. Mir, "Scalability, Consistency,
              Reliability and Security in SDN Controllers: A Survey of
              Diverse SDN Controllers", 2021.  Springer Journal of
              Network and Systems Management, 29(2)

   [Ake24]    Akem, A. T. J., Fraysse, G., and M. Fiore, "Encrypted
              Traffic Classification at Line Rate in Programmable
              Switches with Machine Learning", 2024.  IEEE/IFIP Network
              Operations and Management Symposium (NOMS)

   [Ami24]    Amin, R. M., Hammer, P., and A. Butz, "Using Machine
              Learning to Improve Interactive Visualizations for Large
              Collected Traffic Detector Data", 2024.  Proceedings of
              the 29th International Conference on Intelligent User
              Interfaces (ITUI'24), ACM

   [Ans22]    Anser, O., François, J., and I. Chrisment, "Auto-tuning of
              Hyper-parameters for Detecting Network Intrusions via
              Meta-learning", 2022.  IEEE/IFIP Network Operations and
              Management Symposium (NOMS) - AnNet workshop

   [Ans25]    Anser, O., François, J., Chrisment, I., and D. Kondo,
              "TATA: Benchmark NIDS Test Sets Assessment and Targeted
              Augmentation", 2025.  ESORICS (European Symposium on
              Research in Computer Security)

   [Ans26]    Anser, O., François, J., and I. Chrisment, "An Empirical
              Study on Configuration Robustness of Unsupervised IDS",
              2026.  IEEE/IFIP Network Operations and Management
              Symposium (NOMS)

François, et al.         Expires 7 January 2027                [Page 36]
Internet-Draft     Coupling AI and network management          July 2026

   [Arz21]    Arzani, B., Hsieh, K., and H. Chen, "Interpretable
              Feedback for AutoML and a Proposal for Domain-customized
              AutoML for Networking", 2021.  ACM Workshop on Hot Topics
              in Networks (HotNets)

   [Att23]    Attaoui, W., Sabir, E., Elbiaze, H., and M. Guizani, "VNF
              and CNF Placement in 5G: Recent Advances and Future
              Trends", 2023.  IEEE Transactions on Network and Service
              Management, 20(4), pp. 4698-4733

   [Bar23]    Barsellotti, L., Marinis, L. D., Cugini, F., and F.
              Paolucci, "FTG-Net: Hierarchical Flow-to-Traffic Graph
              Neural Network for DDoS Attack Detection", 2023.  IEEE
              International Conference on High Performance Switching and
              Routing (HPSR)

   [Beg19]    Bega, D., Gramaglia, M., Fiore, M., Banchs, A., and X.
              Costa-Perez, "DeepCog: Cognitive Network Management in
              Sliced 5G Networks with Deep Learning", 2019.  IEEE
              INFOCOM

   [Bis23]    Bischl, B., Binder, M., Lang, M., Pielok, T., Richter, J.,
              Coors, S., Thomas, J., Ullmann, T., Becker, M.,
              Boulesteix, A.-L., Deng, D., and M. Lindauer,
              "Hyperparameter optimization: Foundations, algorithms,
              best practices, and open challenges", 2023.  Wiley
              Interdisciplinary Reviews: Data Mining and Knowledge
              Discovery, 13(2)

   [Bos13]    Bosshart, P., Gibb, G., Kim, H.-S., Varghese, G., McKeown,
              N., Izzard, M., Mujica, F., and M. Horowitz, "Forwarding
              metamorphosis: Fast programmable match-action processing
              in hardware for SDN", 2013.  ACM SIGCOMM

   [Bos14]    Bosshart, P., Daly, D., Gibb, G., Izzard, M., McKeown, N.,
              Rexford, J., Schlesinger, C., Talayco, D., Vahdat, A.,
              Varghese, G., and D. Walker, "P4: Programming Protocol-
              Independent Packet Processors", 2014.  SIGCOMM Comput.
              Commun.  Rev. 44

   [Bou18]    Boutaba, R., Salahuddin, M. A., Limam, N., Ayoubi, S.,
              Shahriar, N., Estrada-Solano, F., and O. M. Caicedo, "A
              comprehensive survey on machine learning for networking:
              evolution, applications and research opportunities", 2018.
              Journal of Internet Services and Applications 9, 16

François, et al.         Expires 7 January 2027                [Page 37]
Internet-Draft     Coupling AI and network management          July 2026

   [Bri19]    Brissaud, P.-O., François, J., Chrisment, I., Cholez, T.,
              and O. Bettan, "Transparent and Service-Agnostic
              Monitoring of Encrypted Web Traffic", 2019.  IEEE
              Transactions on Network and Service Management, 16 (3)

   [Cha18]    Chaignon, P., Lazri, K., François, J., Delmas, T., and O.
              Festor, "Oko: Extending Open vSwitch with Stateful
              Filters", 2018.  ACM Symposium on SDN Research (SOSR)

   [Che19]    Chen, Y., Yen, L., Wang, W., Chuang, C., Liu, Y., and C.
              Tseng, "P4-Enabled Bandwidth Management", 2019.  Asia-
              Pacific Network Operations and Management Symposium
              (APNOMS)

   [Che22]    Chen, H., Huang, S., Zhang, D., Xiao, M., Skoglund, M.,
              and H. V. Poor, "Federated Learning Over Wireless IoT
              Networks With Optimized Communication and Resources",
              2022.  IEEE Internet of Things Journal, 9(17)

   [Col22]    Collet, A., Banchs, A., and M. Fiore, "Using Machine
              Learning to Improve Interactive Visualizations for Large
              Collected Traffic Detector Data", 2022.  IEEE Conference
              on Computer Communications (Infocom)

   [Cor22]    Coronado, E., Behravesh, R., Subramanya, T., Fernàndez-
              Fernàndez, A., Siddiqui, M. S., and X. Costa-Pérez, "Zero
              Touch Management: A Survey of Network Automation Solutions
              for 5G and 6G Networks", 2022.  IEEE Communications
              Surveys & Tutorials, 24(4)

   [czb20]    Clemm, A., Zhani, M. F., and R. Boutaba, "Network
              Management 2030: Operations and Control of Network 2030
              Services", 2020.  Springer Journal of Network and Systems
              Management (JNSM)

   [Das23]    Dasari, A.K., Biswas, S. K., Thounaojam, D. M., Devi, D.,
              and B. Purkayastha, "Ensemble Learning Techniques and
              Their Applications: An Overview", 2023.  Advances in
              Cognitive Science and Communications

   [Dat18]    Datta, R., Choi, S., Chowdhary, A., and Y. Park,,
              "P4Guard: Designing P4 Based Firewall", 2018.  IEEE
              Military Communications Conference (MILCOM)

   [Dij19]    Dijkhuizen, N. V., Ham, J. V. D., and X. Li, "A Survey of
              Network Traffic Anonymisation Techniques and
              Implementations", 2014.  ACM Comput.  Surv. 51, 3, Article
              52

François, et al.         Expires 7 January 2027                [Page 38]
Internet-Draft     Coupling AI and network management          July 2026

   [Dub24]    Dube, R., "Faulty use of the CIC-IDS 2017 dataset in
              information security research", 2024.  Journal of Computer
              Virology and Hacking Techniques

   [Dze23]    Dzeparoska, K., Lin, J., Tizghadam, A., and A. Leon-
              Garcia,, "LLM-Based Policy Generation for Intent-Based
              Management of Applications", 2023.  International
              Conference on Network and Service Management (CNSM)

   [Evr19]    Evrard, L., François, J., Colin, J.-N., and F. Beck,
              "port2dist: Semantic Port Distances for Network
              Analytics", 2019.  IFIP/IEEE Symposium on Integrated
              Network and Service Management (IM)

   [For23]    Foroughi, P., Brockners, F., and J.-L. Rougier, "ADT: AI-
              Driven network Telemetry processing on routers", 2023.
              Elsevier Computer Networks (220)

   [Fre15]    Fredrikson, M., Jha, M. S., and T. Ristenpart, "Model
              Inversion Attacks that Exploit Confidence Information and
              Basic Countermeasures", 2015.  ACM SIGSAC Conference on
              Computer and Communications Security

   [Fua24]    Fuad, A., Ahmed, A. H., Riegler, M. A., and T. Čičić, "An
              Intent-based Networks Framework based on Large Language
              Models", 2024.  IEEE International Conference on Network
              Softwarization (NetSoft)

   [Gin24]    Ginige, Y., Dahanayaka, T., and S. Seneviratne,
              "TrafficGPT: An LLM Approach for Open-Set Encrypted
              Traffic Classification", 2024.  Asian Internet Engineering
              Conference

   [gnn1]     Battaglia, P. W. and E. al, "Relational inductive biases,
              deep learning, and graph networks", 2018.  arXiv preprint
              arXiv:1806.01261

   [gnn2]     Rusek, K., Suárez-Varela, J., Mestres, A., Barlet-Ros, P.,
              and A. Cabellos-Aparicio, "Unveiling the potential of
              Graph Neural Networks for network modeling and
              optimization in SDN", 2019.  ACM Symposium on SDN Research

   [Gre19]    Greeshma, K. V. and K. Sreekumar, "Fashion-MNIST
              classification based on HOG feature descriptor using SVM",
              2019.  Int. J.  Innov.  Technol.  Explor.  Eng. 8, 5
              (2019)

François, et al.         Expires 7 January 2027                [Page 39]
Internet-Draft     Coupling AI and network management          July 2026

   [Gup18]    Gupta, A., Harrison, R., Canini, M., Feamster, N.,
              Rexford, J., and W. Willinger, "Sonata: query-driven
              streaming network telemetry", 2018.  ACM SIGCOMM
              Conference

   [Ham21]    Hamdan, M., Hassan, E., Abdelaziz, A., Mohammed, B., Khan,
              S., Vasilakos, A. V., and M.N. Marsono, "Adversarial
              Machine Learning for Network Intrusion Detection Systems:
              A Comprehensive Survey", 2021.  Journal of Network and
              Computer Applications, Elsevier

   [Ham23]    Hammi, S., Hammami, S. M., and L.H. Belguith, "Advancing
              aspect-based sentiment analysis with a novel architecture
              combining deep learning models CNN and bi-RNN with the
              machine learning model SVM", 2023.  Soc. Netw.  Anal.
              Min., 13(117)

   [Hau23]    Hauser, F., Häberle, M., Merling, D., Lindner, S.,
              Gurevich, V., Zeiger, F., Frank, R., and M. Menth, "A
              survey on data plane programming with P4: Fundamentals,
              advances, and applied research", 2023.  Elsevier Journal
              of Network and Computer Applications (212)

   [Hir15]    Hirth, M., Hossfeld, T., Mellia, M., Schwartz, C., and F.
              Lehrieder, "Crowdsourced network measurements: Benefits
              and best practices", 2015.  Computer Networks. 90

   [Hoa21]    Hoang, N. P., Niaki, A. A., Gill, P., and M.
              Polychronakis, "Domain name encryption is not enough:
              Privacy leakage via IP-based website fingerprinting",
              2021.  Proceedings on Privacy Enhancing Technologies

   [Hoo18]    Hooft, J. V. D., Claeys, M., Bouten, N., Wauters, T.,
              Schönwälder, J., Stiller, A. P. B., Charalambides, M.,
              Badonnel, R., Serrat, J., Santos, C. R. P. D., and F. D.
              Turck, "Updated Taxonomy for the Network and Service
              Management Research Field", 2018.  Journal of Network
              System Management (JNSM) 26, 790–808

   [Hua19]    Huang, C., Zhai, S., Talbott, W., Bautista, M. A., Sun,
              S.-Y., Guestrin, C., and J. Susskind, "Addressing the
              Loss-Metric Mismatch with Adaptive Loss Alignment", 2020.
              ICRL

   [Hua25]    Huang, Y., Du, H., Zhang, X., Niyato, D., Kang, J., and Z.
              Xiong, "Large Language Models for Networking:
              Applications, Enabling Techniques, and Challenges", 2025.
              IEEE Network, 39(1)

François, et al.         Expires 7 January 2027                [Page 40]
Internet-Draft     Coupling AI and network management          July 2026

   [Hui22]    Hui, S., Wang, H., Wang, Z., Yang, X., Liu, Z., Jin, D.,
              and Y. Li, "Knowledge Enhanced GAN for IoT Traffic
              Generation", 2022.  ACM Web Conference 2022 (WWW)

   [Jia24]    Jiang, X., Liu, S., Gember-Jacobson, A., Bhagoji, A. N.,
              Schmitt, P., Bronzino, F., and N. Feamster.,
              "NetDiffusion: Network Data Augmentation Through Protocol-
              Constrained Traffic Generation", 2024.  ACM on Measurement
              and Analysis of Computing Systems

   [Jie22]    Jiang, W., "Graph-based deep learning for communication
              networks: A survey", 2022.  Computer Communications,
              Elsevier, vol. 185

   [Jmi22]    Jmila, H. and M. I. Khedher, "Adversarial machine learning
              for network intrusion detection: A comparative study",
              2022.  Computer Networks

   [Jos21]    Jose, M., Lazri, K., François, J., and O. Festor, "InREC:
              In-network REal Number Computation", 2021.  IFIP/IEEE
              International Symposium on Integrated Network Management
              (IM)

   [Jos22]    Jose, M., Lazri, K., François, J., and O. Festor, "NetREC
              Network-wide in-network REal-value Computation.", 2022.
              IEEE International Conference on Network Softwarization
              (NetSoft)

   [Kaf19]    Kafle, V. P., Martinez-Julia, P., and T. Miyazawa,
              "Automation of 5G Network Slice Control Functions with
              Machine Learning", 2019.  IEEE Communications Standards
              Magazine, vol. 3, no. 3, pp. 54-62

   [Ke23]     Ke, H., Dongseong, K. D., and M. R. Asghar, "Adversarial
              Machine Learning for Network Intrusion Detection Systems:
              A Comprehensive Survey", 2023.  IEEE Communications
              Surveys & Tutorials, 25(1), 538-566.

   [Kr14]     Kreutz, D., Ramos, F. M., Verissimo, P. E., Rothenberg, C.
              E., Azodolmolky, S., and S. Uhlig, "Software-defined
              networking: A comprehensive survey", 2015.  Proceedings of
              the IEEE, vol. 103, no. 1, pp. 14-76

   [Lee21]    Lee, K., Smith, L., and P. Abbeel, "Feedback-efficient
              interactive reinforcement learning via relabeling
              experience and unsupervised pre-training", 2021.  arXiv
              preprint arXiv:2106.05091

François, et al.         Expires 7 January 2027                [Page 41]
Internet-Draft     Coupling AI and network management          July 2026

   [Li19]     Li, R., Makhijani, K., Yousefi, H., Westphal, C., Dong,
              L., Wauters, T., and F. D. Turck., "A Framework for
              Qualitative Communications Using Big Packet Protocol",
              2019.  ACM SIGCOMM Workshop on Networking for Emerging
              Applications and Technologies (NEAT)

   [Lin21]    Lin, Z., Bi, S., and Y. -J. A. Zhang, "Optimizing AI
              Service Placement and Resource Allocation in Mobile Edge
              Intelligence Systems", 2021.  IEEE Transactions on
              Wireless Communications, 20(11)

   [Liu16]    Liu, Z., Manousis, A., Vorsanger, G., Sekar, V., and V.
              Braverman, "One Sketch to Rule Them All: Rethinking
              Network Flow Monitoring with UnivMon", 2016.  ACM SIGCOMM
              Conference

   [Liu21]    Liu, Q., Choi, N., and T. Han, "Constraint-Aware Deep
              Reinforcement Learning for End-to-End Resource
              Orchestration in Mobile Networks", 2021.  IEEE
              International Conference on Network Protocols (ICNP)

   [Liu22]    Liu, B., Ding, M., Shaham, S., Rahayu, W., Farokhi, F.,
              and Z. Lin, "When Machine Learning Meets Privacy: A Survey
              and Outlook", 2022.  ACM Comput.  Surv., 54(2)

   [Liu22b]   Liu, L., Engelen, G., Lynar, T., Essam, D., and W. Joosen,
              "Error Prevalence in NIDS datasets: A Case Study on CIC-
              IDS-2017 and CSE-CIC-IDS-2018", 2022.  IEEE Conference on
              Communications and Network Security (CNS)

   [Lop20]    López, J., Labonne, M., Poletti, C., and D. Belabed,
              "Priority Flow Admission and Routing in SDN: Exact and
              Heuristic Approaches", 2020.  IEEE International Symposium
              on Network Computing and Applications (NCA)

   [Mac22]    Machado, I. A., Costa, C., and M. Y. Santos, "Data Mesh:
              Concepts and Principles of a Paradigm Shift in Data
              Architectures", 2022.  Procedia Computer Science, Volume
              196

   [Mar18]    Martinez-Julia, P., Kafle, V. P., and H. Harai,
              "Exploiting External Events for Resource Adaptation in
              Virtual Computer and Network Systems", 2018.  IEEE
              Transactions on Network and Service Management, Vol. 15,
              N. 2,

François, et al.         Expires 7 January 2027                [Page 42]
Internet-Draft     Coupling AI and network management          July 2026

   [Mar20]    Martinez-Julia, P., Kafle, V. P., and H. Asaeda,
              "Explained Intelligent Management Decisions in Virtual
              Networks and Network Slices", 2020.  Conference on
              Innovation in Clouds, Internet and Networks and Workshops
              (ICIN)

   [Mcc93]    McCanne, S. and V. Jacobson, "The BSD packet filter: a new
              architecture for user-level packet capture", 1993.  USENIX
              Winter Conference

   [Mek24]    Mekrache, A., Ksentini, A., and C. Verikoukis, "Intent-
              Based Management of Next-Generation Networks: an LLM-
              Centric Approach", 2024.  IEEE Network, 38(5)

   [Mor18]    Moritz, P., Nishihara, R., Wang, S., Tumanov, A., Liaw,
              R., Liang, E., Elibol, M., Yang, Z., Paul, W., Jordan, M.,
              and I. Stoica, "Ray: A Distributed Framework for Emerging
              AI Applications", 2018.  USENIX Symposium on Operating
              Systems Design and Implementation (OSDI)

   [Mus18]    Musumeci, F., Rottondi, C., Nag, A., Macaluso, I., Zibar,
              D., Ruffini, M., and M. Tornatore, "An overview on
              application of machine learning techniques in optical
              networks", 2018.  IEEE Communications Surveys & Tutorials,
              21(2), 1383-1408.

   [Mut24]    Muteba, F., Dahanayaka, T., and S. Seneviratne, "Digital
              twin (DT)-based predictive maintenance of a 6G
              communication network", 2024.  Procedia Computer Science,
              Elsevier, vol. 238

   [Naj24]    Najm, I. A., Saeed, A. H., Ahmad, B.A., Ahmed, S. R.,
              Sekhar, R., Shah, P., and B.S. Veena, "Enhanced Network
              Traffic Classification with Machine Learning Algorithms",
              2024.  Proceedings of the Cognitive Models and Artificial
              Intelligence Conference

   [Nas21]    Nasr, M., Bahramali, A., and A. Houmansadr, "Defeating
              DNN-Based Traffic Analysis Systems inReal-Time With Blind
              Adversarial Perturbations", 2021.  USENIX Security
              Symposium

   [Ngu20]    Nguyen, T. G., Phan, T. V., Hoang, D. T., Nguyen, T. N.,
              and C. So-In, "Efficient SDN-based traffic monitoring in
              IoT networks with double deep Q-network", 2020.
              International conference on computational data and social
              networks, Springer

François, et al.         Expires 7 January 2027                [Page 43]
Internet-Draft     Coupling AI and network management          July 2026

   [Puj21]    Pujol-Perich, D., Suárez-Varela, J., Xiao, S., Wu, B.,
              Cabello, A., and P. Barlet-Ros, "NetXplain: Real-time
              explainability of Graph Neural Networks applied to
              Computer Networks", 2021.  MLSys workshop on Graph Neural
              Networks and Systems (GNNSys)

   [Raj24]    Rajasekaran, S., Ghobadi, M., and A. Akella, "CASSINI:
              Network-Aware Job Scheduling in Machine Learning
              Clusters", 2024.  USENIX Symposium on Networked Systems
              Design and Implementation (NSDI)

   [Rex06]    Rexford, J., "Route optimization in IP networks", 2006.
              Handbook of Optimization in Telecommunications (pp.
              679-700), Springer

   [Rin17]    Ring, M., Dallmann, A., Landes, D., and A. Hotho, "IP2Vec:
              Learning Similarities Between IP Addresses", 2017.  IEEE
              International Conference on Data Mining Workshops (ICDMW)

   [Sah23]    Saha, N., Zangooei, M., Golkarifard, M., and R. Boutaba,
              "Deep Reinforcement Learning Approaches to Network Slice
              Scaling and Placement: A Survey", 2023.  IEEE
              Communications Magazine, 61(2)

   [Sal20]    Salman, O., Elhajj, I. H., Kayssi, A., and A. Chehab, "A
              review on machine learning–based approaches for Internet
              traffic classification", 2020.  Springer Annals of
              Telecommunication

   [Sch21]    Schneider, S., Qarawlus, H., and H. Karl, "Distributed
              Online Service Coordination Using Deep Reinforcement
              Learning", 2021.  IEEE International Conference on
              Distributed Computing Systems (ICDCS)

   [Sco11]    Coull, S. E., Monrose, F., and M. Bailey, "On Measuring
              the Similarity of Network Hosts: Pitfalls, New Metrics,
              and Empirical Analyses", 2011.  NDSS

   [Sen04]    Sen, S., Spatscheck, O., and D. Wang, "Accurate, scalable
              in-network identification of p2p traffic using application
              signatures", 2004.  ACM International conference on World
              Wide Web (WWW)

   [Sen24]    Senevirathna, T., La, V. H., Marchal, S., Siniarski, B.,
              Liyanage, M., and S. Wang, "A Survey on XAI for 5G and
              Beyond Security: Technical Aspects, Challenges and
              Research Directions", 2024.  IEEE Communications Surveys &
              Tutorials

François, et al.         Expires 7 January 2027                [Page 44]
Internet-Draft     Coupling AI and network management          July 2026

   [Sha21]    Shapira, T. and Y. Shavitt, "FlowPic: A Generic
              Representation for Encrypted Traffic Classification and
              Applications Identification", 2021.  IEEE Transactions on
              Network and Service Management, 18(2), pp. 1218-1232

   [Sha21b]   Shahraki, A., Abbasi, M., Taherkordi, A., and A. D.
              Jurcut,, "A comparative study on online machine learning
              techniques for network traffic streams analysis", 2022.
              Elsevier Computer Networks (207)

   [Sha24]    Latif, S., Boulila, W., Koubaa, A., Zou, Z., and J. Ahmad,
              "DTL-IDS: An optimized Intrusion Detection Framework using
              Deep Transfer Learning and Genetic Algorithm", 2024.
              Elsevier Journal of Network and Computer Applications

   [Sha25]    Sharma, G. and B. Priya, "Chapter 26 - Future now:
              unleashing the potentials of machine learning algorithms
              in 6G", 2025.  Advances in Computational Methods and
              Modeling for Science and Engineering, Morgan Kaufmann,
              331-339

   [Sil16]    Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre,
              L., Driessche, G. V. D., Schrittwieser, J., Antonoglou,
              I., Panneershelvam, V., Lanctot, M., Dieleman, S., Grewe,
              D., Nham, J., Silver, D., Sutskever, I., Lillicrap, T.,
              Leach, M., Kavukcuoglu, K., Graepel, T., and D. Hassabis,
              "Mastering the game of Go with deep neural networks and
              tree search", 2016.  Nature, vol. 529 (2016), pp. 484-503

   [Sin22]    Singh, A., Prakash, S., and S. Singh, "Optimization of
              reinforcement routing for wireless mesh network using
              machine learning and high-performance computing", 2022.
              Concurrency Computat Pract Exper., 35(15)

   [Sol20]    Soliman, H. M., Salmon, G., Sovilij, D., and M. Rao, "A
              Graph Neural Network Approach for Scalable and Dynamic IP
              Similarity in Enterprise Networks", 2020.  IEEE
              International Conference on Cloud Networking (CloudNet)

   [Ste92]    Stern, D. and P. Chemouil, "A Diagnosis Expert System for
              Network Traffic Management", 1992.  Networks, Kobe, Japan

   [Sul23]    Sulaiman, M., Moayyedi, A., Ahmadi, M., Salahuddin, M. A.,
              Boutaba, R., and A. Saleh, "Coordinated Slicing and
              Admission Control Using Multi-Agent Deep Reinforcement
              Learning", 2023.  IEEE Transactions on Network and Service
              Management, 20(2), pp. 1110-1124

François, et al.         Expires 7 January 2027                [Page 45]
Internet-Draft     Coupling AI and network management          July 2026

   [Swa23]    Swamy, T., Zulfiqar, A., Nardi, L., Shahbaz, M., and K.
              Olukotun, "Homunculus: Auto-Generating Efficient Data-
              Plane ML Pipelines for Datacenter Networks", 2023.  ACM
              International Conference on Architectural Support for
              Programming Languages and Operating Systems (ASPLOS)

   [Tan20]    Tangari, G., Charalambides, M., Pavlou, G., Grazian, C.,
              and D. Tuncer, "Classification-assisted Query Processing
              for Network Telemetry", 2020.  Network Traffic Measurement
              and Analysis Conference (TMA)

   [Tan20b]   Lizhuang, T., Wei, S., Zhenyi, Z., Jingying, M., Xiaoxi,
              L., and L. Na, "In-band Network Telemetry: A Survey",
              2020.  Computer Networks. 186. 10.1016

   [Tan21]    Tang, F., Mao, B., Kato, N., and G. Gu, "Comprehensive
              Survey on Machine Learning in Vehicular Network:
              Technology, Applications and Challenges", 2023.  IEEE
              Communications Surveys & Tutorials, 23(3), 2027-2057.

   [tl1]      Torrey, L. and J. Shavlik, "Transfer learning", 2010.
              Handbook of research on machine learning applications and
              trends: algorithms, methods, and techniques

   [Urb23]    Urbanowicz, R., Zhang, R., Cui, Y., and P. Suri,
              "STREAMLINE: A Simple, Transparent, End-To-End Automated
              Machine Learning Pipeline Facilitating Data Analysis and
              Algorithm Comparison", 2023.  Genetic Programming Theory
              and Practice XIX.  Genetic and Evolutionary Computation.
              Springer

   [Val17]    A., V., M., S., D., S., and T. A., "Learning to route",
              2017.  ACM HotNets

   [Vin21]    Vincenzi, M., Lopez-Aguilera, E., and E. Garcia-Villegas,
              "Timely admission control for network slicing in 5G with
              machine learning", 2021.  IEEE Access

   [Wan22]    Wang, J., Han, H., Li, H., He, S., Sharma, P. K., and L.
              Chen, "Multiple Strategies Differential Privacy on Sparse
              Tensor Factorization for Network Traffic Analysis in 5G",
              2022.  IEEE Transactions on Industrial Informatics, 18(3)

   [Wan24]    Wang, X., Yuan, Q., Wang, Y., Gou, G., Gu, C., Yu, G., and
              G. Xiong, "Combine intra- and inter-flow: A multimodal
              encrypted traffic classification model driven by diverse
              features", 2024.  Computer Networks, Elsevier, vol. 245

François, et al.         Expires 7 January 2027                [Page 46]
Internet-Draft     Coupling AI and network management          July 2026

   [Wu21]     Wu, Y., Zhang, K., and Y. Zhang, "Digital Twin Networks: A
              Survey", 2021.  IEEE Internet of Things Journal, 8(18)

   [Wu22]     Wu, X., Xiao, L., Sun, Y., Zhang, J., Ma, T., and L. He,
              "A survey of human-in-the-loop for machine learning",
              2022.  Elsevier Future Generation Computer Systems

   [Wu24]     Wu, D., Wang, X., Qiao, Y., Wang, Z., Jiang, J., Cui, S.,
              and F. Wang, "NetLLM: Adapting Large Language Models for
              Networking", 2024.  ACM SIGCOMM 2024 Conference

   [XAI]      Samek, W., Wiegand, T., and K.-R. Müller, "Explainable
              artificial intelligence: Understanding, visualizing and
              interpreting deep learning models", 2017.  arXiv preprint
              arXiv:1708.08296

   [Xie18]    Xie, J., Yu, F. R., Huang, T., Xie, R., Liu, J., Wang, C.,
              and Y. Liu, "A survey of machine learning techniques
              applied to software defined networking (SDN): Research
              issues and challenges", 2018.  IEEE Communications Surveys
              & Tutorials

   [Xin24]    Xing, J., Hsu, K.-F., Xia, Y., Cai, Y., Li, Y., Zhang, Y.,
              and A. Chen, "Occam: A Programming System for Reliable
              Network Management", 2024.  ACM European Conference on
              Computer Systems (Eurosys)

   [Xu18]     Z., X., J., T., J., M., W., Z., Y., W., H., L. C., and Y.
              D., "Experience-driven networking: A deep reinforcement
              learning based approach", 2018.  IEEE INFOCOM

   [Yan18]    Yang, T., Jiang, J., Liu, P., Huang, Q., Gong, J., Zhou,
              Y., Miao, R., Li, X., and S. Uhlig, "Elastic sketch:
              adaptive and fast network-wide measurements", 2018.  ACM
              SIGCOMM Conference

   [Yan20]    Yang, H., Alphones, A., Xiong, Z., Niyato, D., Zhao, J.,
              and K. Wu,, "Artificial-Intelligence-Enabled Intelligent
              6G Networks", 2020.  IEEE Network, vol. 34, no. 6, pp.
              272-280

   [Yu14]     Yu, Y., Qian, C., and X. Li, "Distributed and
              collaborative traffic monitoring in software defined
              networks", 2014.  ACM Hot topics in software defined
              networking

François, et al.         Expires 7 January 2027                [Page 47]
Internet-Draft     Coupling AI and network management          July 2026

   [Zha21]    Zhao, J., Jing, X., Yan, Z., and W. Pedrycz, "Network
              traffic classification for data fusion: A survey", 2021.
              Information Fusion, Elsevier, vol. 72

   [Zha24]    Zhang, K., Samaan, N., and A. Karmouch, "A Machine
              Learning-Based Toolbox for P4 Programmable Data-Planes",
              2024.  IEEE Transactions on Network and Service
              Management, 21(4), pp. 4450-4465

   [Zil20]    Meng, Z., Wang, M., Bai, J., Xu, M., Mao, H., and H. Hu,
              "Interpreting Deep Learning-Based Networking Systems",
              2020.  ACM SIGCOMM

Acknowledgments

   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

   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 the Bank of Canada's
   Governing Council.

Authors' Addresses

   Jérôme François
   University of Luxembourg and Inria
   6 Rue Richard Coudenhove-Kalergi
   L- Luxembourg
   Luxembourg
   Email: jerome.francois@uni.lu

   Alexander Clemm
   Independent
   United States of America
   Email: ludwig@clemm.org

François, et al.         Expires 7 January 2027                [Page 48]
Internet-Draft     Coupling AI and network management          July 2026

   Dimitri Papadimitriou
   3NLab Belgium Research Center
   Leuven
   Belgium
   Email: papadimitriou.dimitri.be@gmail.com

   Stenio Fernandes
   Canada Post
   Canada
   Email: stenio.fernandes@ieee.org

   Stefan Schneider
   Digital Railway (DSD) at Deutsche Bahn
   Germany
   Email: stefanbschneider@outlook.com

François, et al.         Expires 7 January 2027                [Page 49]