Reliability Analysis of Random Telegraph Noisebased True Random Number Generators

The diffusion of Internet of Things (IoT) and edge computing poses several security challenges and demands the development of security primitives implemented at the hardware level. Among the technologies that are being considered for applications, true random number generators (TRNGs) exploiting random telegraph noise (RTN) from nanoelectronics devices represent a promising solution thanks to their low cost and high energy efficiency. RTN signal instabilities (e.g., DC drift, temporary inhibition) can potentially affect the TRNG reliability, but a circuit-level reliability analysis accounting for RTN signal instabilities of a RTN-based TRNG is still missing. In this work, we exploit experimental RTN data collected from commonly used gate dielectrics including silicon dioxide (SiO
2
), hafnium dioxide (HfO
2
) and crystalline hexagonal boron nitride (h-BN) and evaluate their performance for TRNG applications in the presence of time variations in RTN signal characteristics and also propose a strategy to enhance the TRNG reliability. Finally, we design and simulate an RTN-based TRNG circuit in a 130 nm CMOS technology and assess its reliability at the circuit-level.