Hardware Trojan Detection Method for Gate-Level Netlists Based on the Idea of Few-Shot Learning
Tong Lü, Zhou Fang, Ning Wu, Fen Ge, Benjun Zhang · 2021 IEEE 21st International Conference on Communication Technology (ICCT) · 2021
In the process of integrated circuits design, third-party companies are usually involved. Therefore, designers cannot fully control the safety of the circuits. Integrated circuits face the potential threat of implanted hardware Trojans (HTs). At present, the detection is mainly to detect well-designed known HTs, and it is difficult to detect unknown new type Trojan. In this letter, we transform the HTs detection into the circuits classification. We adopt the idea of few-shot learning to detect and classify the gate-level netlists features extracted from the Trojan and normal circuits. Compared with other HT detection methods, we only collected 6 different gate-level netlist features of more than 50 circuits to form the training set, and the true positive rate (TPR) reached 70.3%, and the F-measure reached 65.1 %.