Master Thesis - AI-Driven Automated Penetration Testing for Automotive ECUs

Design and Evaluation of an AI-Assisted Penetration Testing Framework for Automotive Electronic Control Units

Introduction 

Background:

Automotive ECUs expose multiple communication interfaces, including Ethernet, DoIP, UDS, and CAN. Security testing often requires manual execution, protocol-specific expertise, and analysis of extensive diagnostic logs.

Objective

The thesis worker will:

• Develop an AI-assisted framework for automated ECU security testing.
• Integrate existing penetration testing tools and test scripts.
• Use LLM-based agents to generate and select security test cases.
• Automatically analyze ECU responses and identify anomalous behavior.
• Compare AI-assisted testing against traditional manual and scripted testing.
  
  

Research questions:

1. How effectively can AI automate automotive ECU penetration testing?
2. Can AI identify vulnerabilities missed by conventional test cases?
3. How much can AI reduce testing time while maintaining accuracy?
Expected deliverables 

A Python-based proof of concept integrated with a simulated ECU or authorized test environment, automated security test generation, and a performance evaluation covering detection rate, false positives, execution time, and coverage.

Value for Scania:

Reduces repetitive testing activities and supports continuous cybersecurity validation across ECU software releases.