SRAM PUFs for Device Authentication on Resource-Constrained Systems

Physically unclonable functions (PUFs) are a vital component of many state-of-the-art techniques to ensure that software is executable only on a predetermined, unique device and becomes unusable on any other. SRAM PUFs are especially suitable for protecting existing hardware used in productive systems, as almost any microcontroller unit (MCU) already has SRAM onboard.
In this paper, we present a novel approach utilizing SRAM PUFs, devised specifically for use on highly resource-constrained devices like MCUs. SRAM-based PUFs exhibit noise due to environmental parameters such as temperature. Therefore we propose a deterministic fuzzy extractor to uniquely identify a device despite the noisy PUF. The memory requirements of our fuzzy extractor rises with noisier data, thus we introduce the concept of stability masks, to preselect the PUF data. Our results show that stability masks significantly improve the reconstruction capabilities of the fuzzy extractor while still functioning with severe memory limitations. We demonstrate our concepts of an automated, scalable multi-stage process implemented in a functional prototype.