Physical Unclonable Functions’ Reliability Modeling for Wireless Sensor Networks

Distributed sensor networks based on wireless communication have been increasingly employed in recent years, but they have become also vulnerable to cyberattacks that can threaten system's security and availability. Therefore, the adoption of techniques to mitigate security risks, such as authentication procedures and cryptographic algorithms, has become of utmost importance, in order to protect sensor networks against cyberattacks. In this regard, Physical Unclonable Functions (PUFs) have been proven to be a lowcost and secure option to store authentication passwords and cryptographic keys. However, PUFs suffer from reliability issues due to variations of temperature and power supply voltage, with a consequent impact on system's availability. The reliability of PUFs is typically expressed by an overall metric, that however does not describe the PUF reliability dependency on the number of bits in the PUF response. Based on these considerations, in this paper we introduce a simple model to evaluate the probability of errors in the PUF response, as a function of the number of bits in the PUF response. The application of our model to a traditional ring oscillator based PUF will show its high accuracy in estimating the error probability with respect to Monte Carlo simulations performed considering statistical variations of temperature and power supply voltage. Our model will show that, as expected, the PUF reliability decreases with the increase in the number of bits in the PUF response which, reversely, enables to increase security. Therefore, our model can be adopted to estimate the optimal number of PUF output bits that enable to achieve the best trade-off between availability and security.