Reconfigurable and Machine Learning Attack Resilient Spintronics based Physically Unclonable Function Design for constrained IoT Systems
Implementing Organization
Dr. Shyama Prasad Mukherjee International Institute Of Information Technology, Chhattisgarh
Principal Investigator
Dr. Deepika Gupta
Dr. Shyama Prasad Mukherjee International Institute Of Information Technology, Chhattisgarh
CO-Principal Investigator
Dr. Japa Aditya
Hyderabad, Telangana-500075
Koneru Lakshmaiah Education Foundation
Project Overview
"The Internet of Things (IoT) devices often face resource constraints, with limited computations, low area, and battery budgets. Current cryptographic key-based authentication protocols are not lightweight enough to address these challenges, and stored cryptographic keys in non-volatile memories are highly vulnerable to invasive attacks. Physically unclonable functions (PUFs) are emerging as a hardware security tool for implementing keyless security strategies. PUFs leverage innate manufacturing variations during IC fabrication and produce unique keys on-the-fly to avoid storing cryptographic keys in non-volatile memory. However, CMOS based PUFs require large post-processing units, large area, and energy consumption overheads. Spintronic devices have attracted attention due to their nonvolatility, 3-D incorporation, and scalability. However, spintronics transfer torque magnetic tunnel junction (STT-MTJ) still suffers from issues like long incubation time, high switching current densities, and read current disturbance. A novel switching mechanism, voltage gated spin orbit torque (SOT), has been developed to address these challenges, improving switching reliability and lower switching energy consumption.
This project aims to design energy-efficient and ultra-lightweight spintronics PUF design for resource-constrained IoT, verifying its vulnerability against side-channel attacks and machine learning-based modeling attacks. The project develops test-setups to derive performance metrics and robustness against power side-channel attacks and machine learning attacks, which can be used for other PUF designs. The designs, processes, and test setups developed can be made available to the research community through a website."