Hardware trojan detection for gate-level ICs using signal correlation based clustering

Abstract

Malicious tampering of the internal circuits of ICs can lead to detrimental results. Insertion of Trojan circuits may change system behavior, cause chip failure or send information to a third party. This paper presents an information-theoretic approach for Trojan detection. It estimates the statistical correlation between the signals in a design, and explores how this estimation can be used in a clustering algorithm to detect the Trojan logic. Compared with the other algorithms, our tool does not require extensive logic analysis. We neither need the circuit to be brought to the triggering state, nor the effect of the Trojan payload to be propagated and observed at the output. Instead we leverage already available simulation data in this information-theoretic approach. We conducted experiments on the TrustHub benchmarks to validate the practical efficacy of this approach. The results show that our tool can detect Trojan logic with up to 100% coverage with low false positive rates.