Information-theoretic analysis of capacitive physical unclonable functions

Physical unclonable functions (PUFs) can be used as a cost-effective means to store cryptographic key material in an unclonable way. In coating PUFs, keys are generated from capacitance measurements of a coating containing many randomly distributed particles with different dielectric constants. We introduce a physical model of coating PUFs by simplifying the capacitance sensors to a parallel plate geometry. We estimate the amount of information that can be extracted from the coating. We show that the inherent entropy is proportional to (log¿n)3/2, where n is the number of particles that fit between the capacitor plates in a straight line. However, measurement noise may severely reduce the amount of information that can actually be extracted in practice. In the noisy regime the number of extractable bits is, in fact, a decreasing function of n. We derive an optimal value for n as a function of the noise amplitude, the PUF geometry, and the dielectric constants.