Large Language Model (LLM) for Networking

• Topic: Large Language Model (LLM) for Networking
• Time: 16:00-17:30 March 20, Wednesday (Brisbane, Australia time)
• Location: Room P6-7
• Host: Yong Cui (Tsinghua University)
• IETF Side meeting wiki: https://wiki.ietf.org/meeting/119/sidemeetings#meeting-wednesday
• Online access method (videoconference): https://ietf.webex.com/meet/ietfsidemeeting2

Motivations

Large language models (LLMs), exemplified by GPT, have achieved remarkable performance in various tasks, such as machine translation, text-to-image generation, and embodied intelligence. Thanks to their vast number of parameters, LLMs can memorize a massive amount of knowledge and utilize tools based on commands. We believe that LLMs can also assist with tasks within networking scenarios. This side meeting will explore the transformative potential of LLMs in the networking domain and open the discussion on potential standards for this topic.

Agenda

• [2 minutes] Opening
• [15 minutes] LLM for Networking: an Overview
  • Speaker: Xiaohui Xie (Tsinghua University)
  • Abstract: The emergence of large language models (LLMs) has brought new solutions to tasks in the field of networking. LLMs can assist networking tasks in providing interpretability, generating configuration strategies, and invoking tools. In this report, we will introduce the background of LLMs, some attempts of applying LLMs in the networking domain, and an attempt to propose an abstract framework. Challenges and future research directions will also be discussed.
  • Slides: slides
• [15 minutes] Using Machine Learning and Word Embedding to Characterise the DDoS landscape with DDoS2Vec
  • Speaker: Ravjot Singh Samra (University of Waikato)
  • Abstract: Volumetric Distributed Denial of Service (DDoS) attacks have been a severe threat to the Internet for more than two decades. Some success in mitigation has been achieved based on numerous defensive techniques created by the research community, implemented by the industry, and deployed by network operators. However, evolution is not a privilege of mitigations, and DDoS attackers have found better strategies and continue to cause harm. A key challenge in winning this race is understanding the various characteristics of DDoS attacks in network traffic at scale and in a realistic manner. In this paper, we propose DDoS2Vec, a novel approach to characterise DDoS attacks in real-world Internet traffic using Natural Language Processing (NLP) techniques. DDoS2Vec is a domain-specific application of Latent Semantic Analysis that learns vector representations of potential DDoS attacks. We look into the link between natural language and computer network communication in a way that has not been previously studied. Our approach is evaluated on a large-scale dataset of flow samples collected from an Internet eXchange Point (IXP) in one year. We evaluate the performance of DDoS2Vec via multi-label classification in a Machine Learning (ML) scenario. DDoS2Vec characterises DDoS attacks more clearly than other baselines - including NLP-based approaches inspired by recent networks research and a basic non-NLP solution.
  • Slides: slides
• [15 minutes] Thinking and Practice: LLM for Cybersecurity
  • Speaker: Linzhe Li (Beijing Zhongguancun Lab)
  • Abstract: Large model technology is an accelerator of the technological revolution in the cybersecurity industry. This presentation will introduce the architecture and key technologies of a large model of the future network security industry. Some practice and the possible applications of LLM in the field of cybersecurity will also be mentioned.
  • Slides: slides
• [15 minutes] Exploiting AI-planning and NLP for Achieving Autonomous Network Reconfiguration
  • Speaker: Angela María Rodríguez (Comfacauca University Corporation)
  • Abstract: Realizing autonomic network reconfiguration is pivotal for achieving self-driving networks. Our approaches ATRAP and NORA introduce a tenant-oriented architecture based on the monitor-analyze-plan-execute-knowledge method (MAPE-K loop) and automated planning (AP) for computing 5G/6G network reconfiguration plans autonomously. NORA uses Natural Language Processing to transform network management policies expressed by the network slice tenants in pure natural language into the goals of an AP-problem (AP-goals). Templates are used to combine such AP-goals with both, the current network status and available reconfiguration actions, and come up with a reconfiguration plan that turns the network from a source -undesired- configuration into a target configuration.
  • Slides: slides
• [15 minutes] Usecases of AI for Network
  • Speaker: Xiaoqiu Zhang (China Mobile)
  • Abstract: The AI-driven network will become an important direction for the future evolution of networks. This presentation will provide a brief overview of some usecases for AI in networks, particularly in the areas of AI for network services, operations and self-learning. Furthermore, we aim to build an AI-driven IP network architecture.
  • Slides: slides
• [20 minutes] Free Discussion (Several open questions will be provided soon)