A side channel based power analysis technique for hardware trojan detection using statistical learning approach

Abstract

Hardware Trojan (HT) is an intentional and the undesired modification of the integrated circuit (IC) and major security issue for the semiconductor industry. HT alters the normal working of IC, can leak the secret information or may damage the IC permanently. Due to the small size of the devices on IC, detection of trojan is very difficult by normal testing methods. In this paper, a side channel based trojan detection technique using power analysis is used to detect the trojan infected IC. Here a trust-hub test bench circuit is used to validate trojan detection technique in which the Trojan is inserted on AES-128 bit crypto core. The trojan detection is improved by analyzing the power of IC without trojan (Golden model) and IC with trojan (Trojan model) and by comparing the mean of power traces of both the IC. Statistical data analysis is performed and statistical parameters of power are calculated which are then used as feature vectors. These feature vectors are reduced by using Principal Component Analysis (PCA) algorithm and then classified using Linear Discriminant Analysis (LDA) which discriminates between the Golden and Trojan model and detects the trojan infected IC from the IC under test with 100% accuracy.