MSc Dissertation: Adversarial Training for Deep-Learning IDS in the Internet of Vehicles
Hardening neural IDS against FGSM and PGD attacks on the CAN bus.
Master's dissertation investigating whether adversarial training can make deep-learning intrusion detection systems robust against evasion attacks on the Controller Area Network (CAN bus), the primary in-vehicle communication backbone. The work uses the CICIoV2024 dataset and compares a 1D-CNN baseline against a Random Forest baseline under Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) attacks, then evaluates a Multi-Strategy Adversarial Training defence implemented via the Adversarial Robustness Toolbox (ART).