CARdea: Two-Phase Plausibility-Based Vehicle-to-Vehicle Anomaly Detection...
summary
A practical V2V anomaly detector called CARdea implements a two-phase pipeline: a lightweight in-vehicle statistical stage that flags anomalies with up to 98% sensitivity in 0.1 ms and an optional ML stage (random forest performs best) run on edge or cloud for ambiguous frames. Evaluation uses 108 hours of simulated BSMs across seven attack types, a Raspberry Pi 3 for phase one, and shows only ~10% of frames offloaded and under 1.8 MB/hour bandwidth while achieving high true positive rates.
AI-written summary · about
chapters (8)
- 0:00CARdea and V2V anomaly detection
- 1:13Threat model and V2V attacks
- 2:36Limitations of existing approaches
- 3:38CARdea two-phase system
- 5:06CARdea deployment options
- 6:28Phase one and two interaction
- 8:14CARdea performance and efficiency
- 10:44Conclusion and future work
related talks
14mStateful Behavior Inference and Runtime Enforcement for Vehicle Network Security
46mConnected Vehicle is far more than just the car
18mModels on the Move: Towards Feasible Embedded AI for Intrusion Detection on...
12mStopping V2X Threats at the Source: Hardware Cybersecurity for Automotive
43mConnected Cars - What could possibly go wrong?
29mAnomaly Detection
10mWIP: From Detection to Explanation: Using LLMs for Adversarial Scenario Analysis...
43mSecuring Software-Defined Vehicles: From Threat Awareness To Holistic Cybersecur...