We organise the team seminars with the talks by newcomers in the team on September 10, 2024 at 2 pm. Seminar location: 2 rue Conté, CNAM, Paris, France.
Speaker Yasmine Chaouche
Title: Anomaly Detection in Automated xG Networks with AI Techniques
Abstract: In softwarized networks, such as 5G and beyond (xG), where the control and data planes are separated, network programmability is key to enhancing flexibility and scalability. However, the increasing complexity and size of these networks, along with the vast amounts of generated data, introduce new challenges, particularly in terms of scalability and security. Anomalies, which are deviations from expected normal behavior, can further complicate these challenges. Some anomalies may be introduced by attackers who exploit vulnerabilities, such as flooding communication channels between the controller and switches, leading to control plane saturation and potential network failure. To mitigate these risks, anomaly detection is essential for identifying deviations from normal behavior and enabling proactive defense mechanisms. Artificial intelligence (AI) offers a powerful and flexible approach for detecting anomalies and intrusions in various network structures. In addition, the inherent complexity and changing nature of xG networks make Federated Learning (FL) an ideal approach to anomaly detection. FL enables distributed learning across multiple devices or nodes, which in turn addresses the challenges posed by dynamic user behavior and the need for sustained performance. The objective of this work is to develop real-time anomaly detection for xG networks, particularly in distributed, low-latency infrastructures. The goal is to design and implement automatic anomaly detection algorithms that align with the architecture and requirements of xG networks, ultimately enhancing network security and resilience.
Bio: Holder of an engineering degree and a master’s in computer systems, data science, and AI. Currently a PhD student working on anomaly detection in automated networks using AI techniques.
Speaker Mehdi Boudjelli
Title: Self-Supervised Anomaly Detection in Time Series
Abstract: The presentation will provide an overview of self-supervised anomaly detection in time series data. It will discuss the different approaches that are most used in current research. Additionally, it will introduce a novel approach that I’ve been working on
Speakers Stefano Secci and Pedro Braconnot Velloso
Title: Feedback on selected SIGCOMM papers
