Hardware Trojan Detection Using Machine Learning

The cyber-physical system’s security depends on the software and underlying hardware. In today’s times, securing hardware is difficult because of the globalization of the Integrated circuit’s manufacturing process. The main attack is to insert a “backdoor” that maliciously alters the original circuit’s behaviour. Such a malicious insertion is called a hardware trojan. In this thesis, the Random Forest Model has proposed for hardware trojan detection and this research focuses on improving the detection accuracy of the Random Forest model. The detection technique used the random forest machine learning model, which was trained by using the power traces of the circuit behaviour. The data required for training was obtained from an extensive database by simulating the circuit behaviours with various input vectors. The machine learning model was then compared with the state-of-art models in terms of accuracy in detecting malicious hardware. Our results show that the Random Forest classifier achieves an accuracy of 99.80 percent with a false positive rate (FPR)of 0.009 and a false negative rate (FNR) of 0.038 when the model is created to detect hardware trojans. Furthermore, our research shows that a trained model takes less training time and can be applied to large and complex datasets