CANdid - An Open-Access Annotated Dataset of Vehicle CAN Bus Traffic
summary
CANdid provides raw CAN packets, synchronized video and GPS, and timed control captures from 10 vehicles across four manufacturers (2012–2018) collected via Raspberry Pi + PiCAN FD over OBD with two USB cameras and roof GPS. Labeled actions let simple scripts isolate signals; analysis shows inconsistent encodings (indicators in different bytes, and steering stored as 12/16/17-bit signed integers). A feed‑forward and a CNN trained on 100‑message windows identify turns with AUC > 0.9, demonstrating practical cross‑vehicle inference.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 3:43Motivation for CANdid dataset
- 6:30Data collection setup
- 9:03Control captures and labeling
- 11:34Analyzing CAN data for maneuvers
- 14:20Future directions and summary
related talks
14mAutoHack: A Physically Verified Multi-Bus CAN Dataset for Intrusion Detection...
54mCantact: An Open Tool for Automative Exploitation
12mOn the Real-World Applicability of Automotive CAN Intrusion Detection Systems
12mDetect Denial of Service on an Open Embedded Automotive Platform
42mCanspy: A Platform for Auditing Can Devices
14mCANBERT+: An Optimized Multi-Layer BERT-Based Intrusion Detection for CAN Bus
50mAdvanced CAN Injection Techniques for Vehicle Networks
18mModels on the Move: Towards Feasible Embedded AI for Intrusion Detection on...