CANGuard: A Spatio-Temporal CNN-GRU-Attention Hybrid Architecture for Intrusion Detection in In-Vehicle CAN Networks

Collection of recent research (Mar 2026) covering applied ML defenses and privacy for cyber-physical and AI systems. Key contributions: CANGuard — a CNN+GRU+attention spatio-temporal model for CAN-bus intrusion detection (CICIoV2024) addressing DoS/spoofing in vehicles; evaluation of AI for network intrusion detection and identification of side‑channel leakage, showing high in-domain accuracy but brittle performance under dataset shift; novel “mimetic deception” IC camouflaging techniques that intentionally mislead reverse‑engineering and DPA analyses; AVDA — an MCP-based framework automating