Low-rate error-correcting coding techniques for SRAM physically unclonable functions

A physically unclonable function (PUF) is a unique circuit on a chip that produces chip-specific outputs, similar to a digital fingerprint, which can be used for identifying the chip or generating unique security keys. PUFs are seen as valuable in securing digital systems like those in the Internet of Things (IoT) due to their ability to generate chip-unique secrets that serve as a foundational "root of trust." One widely used PUF variant is the SRAM PUF, which leverages the natural characteristics of SRAM components found in most digital chips, making it a practical choice for broad deployment. However, PUFs can exhibit variations between measurements due to noise and bias, affecting their reliability. To address this, secure PUF secret generation often employs a fuzzy commitment scheme (FCS), which incorporates error-correcting codes to correct for these variations, thereby ensuring robust and secure secret generation.

The CORTECH-4-PUF project, in collaboration with Synopsys and the Technical University of Munich (TUM), aims to advance the state of the art in error-correcting codes for PUF-based systems. Existing solutions rely on simple concatenated algebraic block codes, but information theory suggests these are suboptimal for PUF requirements. The project focuses on developing new, modern coding schemes that utilize soft-decision-based probabilistic decoding strategies to better manage the low signal-to-noise ratios, block-length constraints, and bias effects unique to PUFs. By optimizing these error-correcting codes for PUF-specific conditions, CORTECH-4-PUF seeks to enhance both the reliability and security of PUF/FCS solutions for future digital systems.