Random Number Generation
The invention consists of a novel approach for true random number generation (TRNG), also capable of achieving a physical unclonable function (PUF) for hardware security. The objective is met by using 2 memristors in a circuit with serial/parallel configuration, where a random bit is generated by the application of a standard pulse across the memristors, and the bit value results from the comparison between the resistances of the 2 memristors during the read operation. The 2-memristor configuration intrinsically guarantees the uniformity of bit values, namely the probabilities of generating either 0 or 1 are both 50%. The generation of random bits can be repeated an arbitrary number of times for TRNG applications, or can be executed only once to generate a hardware signature of a circuit for PUF applications (e.g., chip security and counterfeits). The unicity of the signature is guaranteed by the intrinsic stochastic variation of the memristor operation.
The 2-memristors layout guarantees the uniformity (50% probability)of generated bits irrespective of the applied voltage. This approach overcomes the limitations of 1-memristor schemes which require instead a pre-calibration of the voltage to ensure equal probability of generated bits irrespective of the applied voltage. This approach overcomes the limitations of 1-memristor schemes which require instead a pre-calibration of the voltage to ensure equal probability of 0 or 1, hence a higher circuit complexity and higher cost.
The invention finds application in systems-on-chip (SoC) requiring generation of random numbers for cryptography application or containing a hardware signature for the purpose of security of the transmitted information. Given the growing amount of systems on chip for internet-of-things (IoT) applications with transmission of sensitive data, there is a significant interest in TRNG and PUF functions with high density, low cost, low power consumption, and meeting the requirements of uniformity, unicity and reliability.