Statistical and Machine Learning-Based Solutions for Trusted and Secure Analog/RF ICs

Unlike the extensive research effort that has been expended over the last decade in understanding the threat of hardware Trojans, piracy, and counterfeiting of digital ICs/IPs, and developing appropriate prevention and detection solutions, the topic remains largely unexplored for their analog/RF counterparts. Given the widespread use of analog functionality (i.e., physical interfaces, sensors, actuators, wireless communications, etc.) in most contemporary systems, a comprehensive understanding of these threats in analog/RF ICs is urgently needed, in order to facilitate the development of pertinent solutions. This presentation will focus on the use of statistics and machine learning towards ensuring security and trust of analog/RF ICs. Specifically, we will first introduce methods for detecting and/or preventing malicious hardware modifications, a.k.a. hardware Trojans, and we will assess their effectiveness using silicon measurements from a custom-designed wireless cryptographic IC. Extensions of these concepts in the wireless networking domain will also be demonstrated using popular WiFi experimentation platforms. We will then review variants of these methods which can be used for distinguishing between genuine and counterfeit ICs, as well as for attesting the manufacturing facility wherein an analog/RF IC was fabricated. Lastly, we will conclude by examining the role that such statistical and machine learning-based methods can play in establishing security and trust in contemporary domains, such as autonomous vehicles and the internet of things.