Title: Enhanced Hardware Trojan detection by Reducing linearity between Features
Program: Master of Science in Electrical and Computer Engineering
Advisor: Dr. Nader Rafla, Electrical and Computer Engineering
Committee Members: Dr. Benjamin Johnson, Electrical and Computer Engineering; and Dr. Jennifer Smith, Electrical and Computer Engineering
Malicious modifications to the design or manufacturing process of an integrated circuit, known as Hardware Trojans, can pose a serious threat to the security, reliability, and functionality of the device, as well as the systems that rely on it. Detecting and mitigating these threats is a critical challenge for the semiconductor industry. Machine Learning has emerged as a promising approach for detecting Hardware Trojans by analyzing the physical or electrical behavior of the device and identifying anomalous or unexpected patterns.
The thesis proposes a method for mitigating the issue of Hardware Trojans through the use of Machine Learning and the reduction of feature linearity by means of Correlation Coefficient. The approach involves feeding the post-synthesis netlist features of the digital hardware design to a machine learning model, and eliminating any interdependence among the features to avoid overfitting of the training data. The study applies the machine learning model to various benchmarks and real-world situations.