Unsupervised IC Security with Machine Learning for Trojan Detection
Conference
Ghimire, A, Amsaad, F, Hoque, T et al. (2023). Unsupervised IC Security with Machine Learning for Trojan Detection
. 20-24. 10.1109/MWSCAS57524.2023.10406045
Ghimire, A, Amsaad, F, Hoque, T et al. (2023). Unsupervised IC Security with Machine Learning for Trojan Detection
. 20-24. 10.1109/MWSCAS57524.2023.10406045
The detection of Hardware Trojans is crucial for ensuring trust in the semiconductor IC supply chain. However, existing detection methods that rely on side-channel analysis often require golden chips for verification. This paper presents a new approach to Hardware Trojan detection that utilizes unsupervised machine learning with side-channel analysis to eliminate the need for golden data. Trojans of varying sizes were implemented on an FPGA to evaluate the method to perform unsupervised clustering and detect anomalies. The proposed model achieved high accuracy, 93%, and improved the detection of small and short-triggered Trojans compared to competing approaches. The unsupervised learning techniques demonstrated a better false positive rate and similar accuracy to supervised approaches such as the KNN classifier, SVM, and Gaussian classifier which require golden data for training. This research contributes a new approach to Hardware Trojan detection that can improve the trustworthiness of semiconductor IC supply chains.
