| Summary: | While much of integrated circuit development over the last few decades has focused on power, performance, and area, hardware security is rapidly gaining prominence as a major consideration during the design process. Particularly, physical side channels that allow reverse engineering of inputs, operation states, and private information must be characterized and protected against. In this thesis, we focus on electromagnetic (EM) physical side channels. EM side channel measurements using a novel quantum diamond microscope are used to protect the integrated circuit supply chain by detecting hardware trojans with high spatial resolution, a wide field of view, and high sensitivity. A hardware trojan detection framework is developed that allows for automated and unbiased detection, using convolutional neural networks instead of principal component analysis for higher accuracy. In addition, the EM side channel is considered as an attack method to gain sensitive data from analog to digital converters (ADC). To this end, we propose a method of protection using conversion randomization that reduces both power and EM side channel leakage with minimal area and accuracy overhead. Due to the long wait times between samples in Internet of Things applications, we are able to make a trade off between conversion time and security. Finally, we consider the use of better routing and placement to augment EM side channel resilience for general digital circuits and capacitive digital to analog converters with no first-order power, area, or performance overhead.
|
|---|