A Secure and Sustainable RISC-V Processor with Intrinsic PUF for Edge AI

The open-source and modular nature of RISC-V has driven its widespread adoption in IoT devices, particularly in edge Artificial Intelligence (AI). However, implementing AI algorithms on the edge results in high power consumption that reduces the battery life of the device. Coupled with the rise in the number of IoT devices, this increases the operational Carbon Footprint (CFP). Another major concern is the impersonation of legitimate IoT devices by adversaries. The adversaries can create clones that impersonate the IoT device and access sensitive data. In this work, we propose an open-source RISC-V ISAbased processor with an approximate multiplier that serves dual purposes, approximate computing and authentication. We reuse the approximate multiplier for Physically-Unclonable-Function (PUF) based authentication. This eliminates the need for separate PUF circuitry, reducing both embodied and operational CFP as compared to state-of-the-art RISC-V-based PUFs. We provide an end-to-end framework that encapsulates the hardware and software modifications required for approximate multiplication and PUF-based authentication within the RISC-V ISA-based processor. Experimental results show 48.24% uniqueness, 47.71% uniformity, and 98.63% reliability against voltage and temperature variations, with up to 55% reduction in embodied CFP and 46% reduction in operational CFP.