Cybersecurity / Machine Learning01
Master's thesis project
OTA-Sentinel
A pre-installation anomaly-screening framework for examining files inside automotive infotainment updates before installation or activation.
- Problem
- Cryptographic verification establishes that an update is authentic and authorized, but does not establish that every accepted file is benign.
- Approach
- Extracted static file characteristics including size, entropy, byte distributions and byte 3-grams, then evaluated multiple unsupervised anomaly-detection models.
- Evaluation
- Tested using vendor Linux-based IVI files, Android-Automotive-derived contents, external malware and controlled file modifications.
Python / TensorFlow / scikit-learn / HDF5 / CUDA


