Enhancing Security of a PUF-based Remote Keyless Entry System using Machine Learning Approach
Zong Tao Augustine Tang, K. Y. Yu, Kii Yuta, Ting-Yu Chen, Arijit Karati · 2024
The Remote Keyless Entry (RKE) system offers users the convenience of remotely locking and unlocking the vehicle, activating the engine through a keychain transmitter. Modern RKE systems are vulnerable to security breaches like the RollJam attack, which exploits signal relaying and interference to activate a vehicle without authorization potentially. This research integrates a “HODOR" RF profiling technique to detect malicious attacks in a lightweight mutual authentication mechanism using Physically Unclonable Functions (PUFs). The enhanced protocol withstands device cloning and RollJam attacks. Experimental results show that the concurrent execution of HODOR and PUFs introduces an overhead of 2.01 seconds, and through training with PUF authentication and machine learning, an accuracy of 99.8% was achieved with a training loss of 2.3%. Integrating our approach into the operational RKE system establishes a device-coupled authentication and malicious behavior detection system, effectively safeguarding various vehicular communication applications.