Collaborative Research: SaTC: CORE: Medium: KIPPER: A Scalable Learning-Guided Hardware IP Protection Platform

University Of Florida, Gainesville FL

Investigators

Abstract

The increasing use of reusable hardware intellectual property (IP) in modern semiconductor design faces significant confidentiality threats, including reverse engineering (RE), IP theft, and piracy throughout its lifecycle. Various active IP protection techniques have been investigated to secure hardware IPs against these threats. With the emergence of a broad range of attacks on these techniques, there is a need for a principled approach to enable robust protection against possible attacks, while achieving low hardware overhead and scalability to large designs. The key novelty of this project is the use of diverse Artificial Intelligence (AI) techniques, such as reinforcement learning and explainable algorithms for automated discovery of new hardware security attacks and deriving commensurate design transformations to effectively protect against these attacks. The research outcomes from this project are used to develop new cybersecurity courses, increase the participation of undergraduate/high-school students in research, and release of open-source tools/datasets for the broader research community. This project also contributes to enhancing the security and trustworthiness of electronic hardware and greatly benefits the semiconductor industry, as well as making a positive impact on national security. This project is developing a new framework for microelectronic security and trust. The approach utilizes a fundamentally different, knowledge-guided systematic design transformation approach for securing IP against RE attacks. This framework mimics the way security researchers and engineers gather knowledge from existing attacks to discover novel attack vectors and their root causes. Furthermore, the method can devise strong defense strategies to mitigate the newly discovered attacks and apply these defense strategies in a scalable manner to create a protected design with minimal design overhead. This project comprises of three parts: (1) the reinforcement learning techniques discover novel attack vectors against logic locking and identify root causes behind the success of these attacks, (2) explainable diverse Artificial Intelligence (AI) is used to extract design transformation rules that can defend against diverse attack vectors, (3) the discovered design rules are applied to successfully meet design and security requirements while also ensuring correctness. The research team is also developing novel security metrics and integrating all the components into a comprehensive IP protection platform for rapid evaluations. Moreover, the research team also extends the framework to other Design-for-Security techniques, such as fault, side-channel, and Trojan attack countermeasures. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.