Project 01 · Automotive Security
A Novel Fingerprint-Based Intrusion Detection Approach to Enhancing In-Vehicle Communication Security
- Category
- Automotive Security
- Duration
- 6 months
- Team
- Group Research
- Supervisor
- Prof. Hafiz Malik
- Institution
- University of Michigan-Dearborn
Modern vehicles have evolved into complex cyber-physical systems with dozens of Electronic Control Units (ECUs) communicating over internal networks such as CAN bus. As connectivity grows, so does the risk of cyber-attacks targeting in-vehicle communication channels. This project, conducted under the guidance of Prof. Hafiz Malik at the University of Michigan-Dearborn, explores an innovative fingerprint-based intrusion detection approach specifically designed to secure such networks against unauthorized access and malicious manipulation.
In this research, I developed a method to capture subtle variations in ECU signal patterns, effectively creating a unique 'fingerprint' for each legitimate ECU. By continuously monitoring real-time traffic and comparing it against stored fingerprints, the system can detect anomalies indicative of spoofed or compromised ECUs. To enhance detection accuracy, I integrated neural network models trained to recognize legitimate versus manipulated signal distortions.
Using Wireshark, I performed ARP spoofing and replay attacks within a controlled lab setup to test detection capabilities. I applied statistical techniques to filter noise and extract meaningful features from raw ECU traffic. By employing time-series analysis and clustering, the system adapts to normal variations without excessive false positives.
This project strengthened my skills in automotive cybersecurity, signal processing, neural networks, and practical ethical hacking in regulated environments. Next, I aim to extend this research with firmware integrity checks and real-time cloud-based threat intelligence.
01
Key features
- ▹ECU Signal Fingerprinting
- ▹Real-time Anomaly Detection
- ▹Neural Network Integration
- ▹Statistical Noise Filtering
- ▹Time-Series Analysis
- ▹Attack Simulation & Testing
02
Challenges
- ▹Capturing subtle signal variations in noisy automotive environments
- ▹Reducing false positives while maintaining high detection accuracy
- ▹Adapting to normal ECU behavior variations over time
- ▹Processing real-time data streams efficiently
03
Outcomes
- ▹Developed novel fingerprinting methodology for ECU identification
- ▹Achieved high detection accuracy for spoofed ECUs
- ▹Demonstrated effectiveness against ARP spoofing and replay attacks
- ▹Gained deep expertise in automotive cybersecurity
Next project