Stateful Behavior Inference and Runtime Enforcement for Vehicle Network Security
summary
Framework infers hierarchical state machines from decoded CAN traces, compiles them to an FPGA checker and combines that with ECU fingerprinting for message authentication. Implementation runs on a Toyota Prius and in the BeamNG.tech simulator; injected-message fuzzing shows state-desynchronization attacks succeed rarely (assertion violations ~0.35% average), pure mode-transition attacks produce none in a 24‑hour sweep, and a history-assisted variant stays under ~0.5%. Enforcement detects injected anomalies, leaves FPGA resources unused, and scales by adding parallel FPGA instances.
AI-written summary · about
chapters (12)
- 0:00Vehicle network security intro
- 1:14Behavior enforcement framework
- 2:37Data for autonomous parking
- 4:02State machine inference
- 5:24Deep spec mining comparison
- 6:05FPGA state machine impl
- 7:48FPGA model scalability
- 8:09Penetration testing & threats
- 9:33Desynchronization attack results
- 10:58Mode transition attacks
- 11:19Scalability & other features
- 12:21Conclusion & future work
related talks
18mModels on the Move: Towards Feasible Embedded AI for Intrusion Detection on...
01h 05mTheory and Security
01mFrom Bugs to Breaches
50mNavigating Chaos: The Consequences of Compromise in Mixed-Autonomy Freeways
12mOn the Real-World Applicability of Automotive CAN Intrusion Detection Systems
34mDriving Safety Forward: Lessons Learned From Deploying OSS in Real-world Automotive
50mAdvanced CAN Injection Techniques for Vehicle Networks
14mAutoHack: A Physically Verified Multi-Bus CAN Dataset for Intrusion Detection...