CIFellows Spotlight highlights the work of the Computing Innovation Fellows (CIFellows) for the computing research community.
Soheil Salehi began his CIFellowship in September 2020 after receiving her PhD from University of Central Florida in May 2020. Soheil is at University of California Davis working with Houman Homayoun, Associate Professor of Electrical and Computer Engineering at University of California Davis.
The remainder of this post is written by Soheil Salehi
My research focus is on applications of AI in secure Internet of Things (IoT) sensing and computing hardware. Currently, I am leading several projects on the topic of AI-enabled security for the IoT supply chain, which takes on a ground-up approach to ensure the reliability, security, and energy efficiency of the IoT hardware. Within this topic, I have tackled many interesting challenges and proposed solutions to eliminate them or mitigate their impacts. In particular, my CIFellows project, called “SHIELD: Secure Hardware for IoT using Emerging-devices against Side-channel Deep-learning attacks,” focuses on addressing the growing challenges of IoT hardware supply-chain security. The proposed approach aims to provide a defense-in-depth mechanism using emerging post-CMOS devices to prevent various attacks, such as netlist reverse engineering attacks and power side-channel attacks, simultaneously.
The current state-of-the-art ASIC industry heavily relies on the globalized fabrication processes and hardware supply-chain model. While this benefits both participants and their global economy, the security of the underlying hardware is compromised due to various emerging hardware security threats such as overproduction, hardware trojan insertion, reverse engineering, IP theft, and counterfeiting. Hardware at the heart of the critical infrastructure consists of various hardware IPs and assets, which their continuous reliable function is deemed necessary to ensure the security of a given nation, its economy, and the public’s health and safety. Additionally, there are no off-the-shelf software or scripts available for deploying the hardware security primitives. To address these challenges, I focused on developing a novel approach that bridges the ongoing research in deep learning and innovative hardware design to increase the security coverage of IoT hardware.
Hardware-oriented attacks are rapidly increasing with the adoption of computing systems in our daily lives and can have significant societal and economic impacts. The results of the project will include deep learning-based attack and mitigating hardware to facilitate broad ranges of secure IoT systems under energy constraints and security demands. The contributions of this project are significantly important to the industry and academia due to the recent hardware security threats in IoT applications. Exciting educational opportunities will be enabled by the development of novel hardware security measures and research of machine learning in the context of security. These novel methodologies will provide new directions for educational initiatives targeting teaching and research activities within the broader academic community. Finally, the broader economic and societal impacts of the proposed work include the feasibility of improved security and energy efficiency towards privacy-preserving IoT hardware.
My technical research interests span Hardware and AI-enabled Security in IoT, Energy-Efficient and Intelligent Signal Conversion and Processing in IoT, Emerging Spin-Based Devices for Low-Power and Reconfigurable Circuits and Architectures, and Neuromorphic and Biologically-inspired AI Hardware.
More information about Soheil’s recent projects and publications is available on his website.