My PhD research project aims to enhance machine learning generalizability and robustness against adversarial attacks in Computer Networks and XIoT (e.g., medical, industrial) environments. This involves developing advanced Network Intrusion Detection Systems (NIDS) and exploring offensive strategies to understand how attackers bypass AI systems, ultimately aiming to develop robust countermeasures for resilience against sophisticated adversarial threats.
The methods employed include integrating state-of-the-art machine learning techniques such as Variational Autoencoders (VAE), Normalizing Flow (NF) models, Reinforcement Learning, Stochastic Shattering Gradient Descent, and Large Language Models (LLMs) in novel ways to improve NIDS robustness and generalizability.
Expected outcomes of my research include the creation of more resilient NIDS capable of maintaining high detection rates and low false positives, even in the presence of sophisticated adversarial threats.