Committee

1. Alenka Zajić (Advisor)

2. Matthieu Bloch (Co-Advisor)

3. David Anderson

4. Mark Davenport

Abstract

The authenticity of integrated circuits is of increasing security and safety concern as more steps in the device manufacturing supply chain are outsourced, especially in light of the current global semiconductor shortages. Integrated circuit validation commonly relies on destructive methods paired with high resolution imaging or automated functional testing, which are costly, time-consuming, or even intractable to detect counterfeit components or stealthy modifications to components. This proposed work takes advantage of the electromagnetic (EM) side channel to remotely capture identifying information from integrated circuits. This research attempts to alleviate the need for time-consuming and destructive hardware validation methods by developing deep learning, active learning, and compressed sensing techniques to robustly detect inauthentic or modified integrated circuits using hyperspectral EM side-channel measurements.