Software lab / virtual CAN security

Software lab

CAN security lab

This software lab models four ECUs on a virtual CAN interface, injects attacks and compares intrusion detection approaches. I implemented the simulated traffic, LSTM-based attack injection, detector baselines and an API/dashboard. The repository is public. These experiments use simulated data; they do not establish performance on an actual N7 Racing vehicle or on its physical PC–STM32 validation bench.

01

Engineering contribution

  • ProblemExplore CAN traffic anomalies and detection on a controllable network.
  • ContributionFour-ECU simulation, attack generation, Isolation Forest/autoencoder/CUSUM detectors and API/dashboard.
  • StackPython, virtual CAN, PyTorch, scikit-learn, FastAPI and Streamlit.
  • LimitVehicle traces and detection latency under real hardware load remain separate validation work.
02

Evidence & scope

Public source