AutoHack: A Physically Verified Multi-Bus CAN Dataset for Intrusion Detection...
summary
A 2023 Hyundai–based dataset captures synchronized C-CAN P-CAN B-CAN traffic with a dual‑channel CAN logger and a single timestamp base, totaling 9.6 billion frames with ~480,000 labeled attack frames. Five live injection types include UDS diagnostic sequences, timing‑opaque replay, ID flooding and short‑burst ID spoofing; physical effects such as engine stalls and emergency call activation are verified, and Random Forest/XGBoost baselines plus AutoHack competition entries consistently fail on replay and spoofing. Data and code are archived on Zenodo.
AI-written summary · about
chapters (13)
- 0:00AutoHack dataset intro
- 1:53Existing CAN dataset limits
- 2:38AutoHack dataset traits
- 3:21Real-time vehicle data
- 4:03Attack injection methods
- 5:29Data organization
- 6:33Dataset features
- 6:54Evaluation setup
- 7:37Spoofing and replay attacks
- 8:44Non-periodic UDS impact
- 9:33Multi-bus training analysis
- 12:57Limitations and future work
- 13:21Q&A
related talks
16mCANdid - An Open-Access Annotated Dataset of Vehicle CAN Bus Traffic
41mIt's easier to break than to patch: a stealthy DoS attack on CAN
36mCar Hacking
37mAutomatic Automotive Hacking
09mLightning Talk: On-Road Driver Identification Dataset with CyberAttack
18mModels on the Move: Towards Feasible Embedded AI for Intrusion Detection on...
50mAdvanced CAN Injection Techniques for Vehicle Networks
58mA Survey of Remote Automotive Attack Surfaces