Car Hacking with Claude Code
The article explores how to use Claude Code, an AI coding assistant from Anthropic, to reverse-engineer CAN bus data from vehicles. It demonstrates using the AI to parse raw CAN logs, identify signals like speed and RPM, and generate Python scripts for decoding, effectively applying large language models to automotive hacking and diagnostics.
Background
- **CAN bus** (Controller Area Network) is the internal communication system used by modern vehicles — a simple two-wire network that connects ECUs (engine, brakes, airbags, etc.) to send short messages. It lacks basic security protections (no encryption, no authentication), making it a prime target for automotive hacking research.
- **CAN bus reverse engineering** normally means manually driving a car, logging traffic, and trying to decode which messages correspond to actions (e.g., pressing the accelerator) — a slow, tedious process.
- **Claude Code** is Anthropic's agentic coding tool (an evolution of the Claude AI assistant) that can autonomously write, edit, and run code to solve tasks.
- This article shows how Claude Code, given a CAN bus interface and a car, can autonomously sniff traffic, correlate messages with actions, and identify the message IDs that control things like throttle, steering, and gear — dramatically automating what was previously a manual expert task.