Statistical randomness of memristive-device-generated sequences

Determine whether random-number sequences derived from the frequency response of the interface-type Au/BiFeO3/Pt/Ti and Au/Nb_xO_y/Al2O3/Nb resistive switching devices are truly random by performing a detailed statistical analysis of their properties, including assessments of unpredictability and randomness consistency across cycles and operating conditions.

Background

The paper proposes using the frequency response of memristive devices as a fingerprint and further suggests leveraging the nonlinear and chaotic behavior observed in the devices' output under periodic inputs to generate random numbers. Specifically, the authors discuss deriving random sequences from the device response of Au/BiFeO3/Pt/Ti (BFO) and Au/Nb_xO_y/Al2O3/Nb (DBMD) devices, highlighting that chaos in the internal state attractors could provide unpredictability.

However, the authors explicitly note that validating the statistical properties—particularly true randomness—of these sequences remains to be done. This open point is crucial for practical hardware security applications, where rigorous randomness evaluation is necessary to ensure robustness and compliance with cryptographic standards.