Critical infrastructure and IoT networks are increasingly managed through edge computing, yet the speed and scale of modern cyberattacks far exceed what human security teams can handle alone.
In this webinar, Dr Tuan Le will explore how reinforcement learning and graph-based AI can empower edge systems to operate defensively and autonomously, detecting threats, making risk-aware decisions, and recovering from compromise without waiting for human intervention.
This session is ideal for researchers, practitioners, and decision-makers working in cybersecurity, AI, critical infrastructure, and IoT systems.
Date: May 15, 2026
Time: 2:00 PM
Location: Zoom: https://newcastleuniversity.zoom.us/j/83866079014
Title: Autonomous Cyber Defence at the Edge – From Risk-Aware Reinforcement Learning to Self-Healing IoT Networks
Abstract: Critical infrastructure and IoT networks are increasingly managed through edge computing, yet the speed and scale of modern cyberattacks far exceed what human security teams can handle alone. This talk will explore how reinforcement learning and graph-based AI can enable edge systems to detect threats, make defensive decisions, and recover autonomously — without waiting for human intervention. I will introduce our experience training intelligent agents to select security controls against multi-stage attacks on industrial networks, and then present a self-healing architecture for IoT environments where compromised devices are automatically isolated, traffic is rerouted, and recovered nodes are safely reintegrated. Overall, these works contribute an end-to-end framework for autonomous threat detection, decision-making, and self-healing at the network edge, with future work targeting hardware deployment and federated learning across distributed edge devices.
About the speaker: Dr Tuan Le is an Assistant Professor at WMG, University of Warwick, where he is part of the Secure Cyber Systems Research Group led by Prof. Carsten Maple. His research focuses on integrating AI with cybersecurity, including reinforcement learning for automated defence, attack-graph-based threat modelling and risk assessment for critical infrastructure systems and IoT.