Research
graph neural networks, intrusion detection, knowledge distillation, CAN bus
My research applies graph neural networks and knowledge distillation to intrusion detection on vehicle CAN bus networks, with a focus on deploying lightweight models on resource-constrained automotive platforms.
Publications & Preprints
KD-GAT: Combining Knowledge Distillation and Graph Attention Transformer for a Controller Area Network Intrusion Detection System
IEEE ITSC 2025 (to appear) — R. Frenken, S.G. Bhatti, H. Zhang, Q. Ahmed
An intrusion detection framework based on Graph Attention Networks and knowledge distillation for CAN bus security. Student model achieves >99.97% accuracy on Car-Hacking and >99.31% on Car-Survival benchmarks. Details · Paper PDF
Multi-Stage Knowledge-Distilled VGAE and GAT for Robust Controller-Area-Network Intrusion Detection
Preprint — R. Frenken, S.G. Bhatti, H. Zhang, Q. Ahmed
A multi-stage intrusion detection system combining variational graph autoencoders and graph attention networks with knowledge distillation for CAN bus security.
CWD-SWGD-IDS: Complementary Fusion of Temporal Context and Structural Graphs for CAN Intrusion Detection via Transformer
IEEE ITSC 2026 (under review) — W.M. Wu, H. Zhang, R. Frenken, Z. Li, S.G. Bhatti, Q. Ahmed
A complementary fusion approach combining temporal context windows and structural graph representations for CAN bus intrusion detection.