Hardware trojan design and detection: a practical evaluation

Hardware Trojan design and detection have been extensively studied during the last years. In this work we investigate non-invasive detection methods utilizing so-called side-channel analysis. In the past, almost all proposed detection techniques have been evaluated based on simulations only and thus, the question remains how well they perform in practice. Therefore, we perform a practical evaluation of two previously published Trojan detection methods based on principal component analysis. We evaluate those methods on various designs of a complete functional lightweight hardware Trojan embedded in a PRESENT block cipher circuit. More precisely, we investigate how well the simulations match our practical results and reveal some shortcomings. Subsequently, we introduce a new detection method exploiting statistical properties of the probability distribution functions built from side-channel measurements and show that it is more robust to measurement noise than previously presented methods.