Lightweight Strategies for Reliability Improvement of PUF-Based Authentication in Resource-Constrained Devices

Cyberattacks represent a serious threat for the security of network-based systems and are responsible for large economic losses every year. In this context, physical unclonable function (PUF)-based authentication can provide access to the network resources to legitimate users only, thus preventing unauthorized accesses. On the other hand, transient disturbances (e.g., noise, temperature and power supply variations) and permanent faults can lead to erroneous PUF responses, resulting in failed authentication and reduced network availability for legitimate users. Error-correcting codes have been proposed in the literature to improve PUF reliability. However, they typically require significant costs in terms of processing power and area overhead, meaning they are often unsuitable for resource-constrained devices, such as low-cost microcontrollers and FPGAs. In this paper, we have investigated strategies based on the use of different kinds of error-detecting and error-correcting codes, as well as their possible combination, with limited requirements in terms of processing power and no need for helper data. These strategies have been evaluated using both a synthetic PUF dataset and a real PUF dataset. The results show that the strategy based on a checksum error-detecting code achieves a good performance in terms of network availability, i.e., an error probability in the order of 10−3 (3.69 × 10−2) when the error on the PUF response (without any ECC) is 12.89% (55.04%), with a low data overhead (1.56% of the PUF challenge size), but it is effective only in the presence of transient disturbances. Instead, the strategy combining the checksum and the Hamming codes provides even higher network availability, i.e., an error probability in the order of 10−4 (1.6 × 10−3) when the error on the PUF response (without any ECC) is 12.89% (55.04%), at the cost of a slightly higher data overhead (7.81% of the PUF challenge size), while also enabling the capability to correct erroneous PUF responses caused by both disturbances and permanent faults.