A novel security technique to generate truly random and highly reliable reconfigurable ROPUF-based cryptographic keys

Abstract

Silicon Physically Unclonable Functions (SPUFs) are delay based PUFs that exploit stochastic manufacturing process variations of Integrated Circuits (ICs) on silicon chips to construct unclonable cryptographic secret keys which are unique for each chip. One variant of SPUFs, named Ring Oscillator (RO) PUFs, is typically used for the authentication of silicon technology devices including FPGA chips. Prior research on the area of ROPUF shows that RO frequencies are affected by spatial systematic process variations and hence the generated responses are not statistically random. In addition to the negative effects of systematic variations on overall ROPUF performance, reduced randomness in the generated responses can lead to major hardware security threats. In this paper, Logarithmic and absolute Diverseness Technique (LDT), a novel security technique based on base-10 logarithm and square root of RO deviations from the global RO mean, is proposed to nullify the effects of spatial systematic variation on the response bits of a unique reconfigurable ROPUF design (r-ROPUF) and improve the reliability of the structure. The proposed technique is implemented on the data obtained from 30 Spartan 3E FPGA chips. IBM-SPSS statistical software is used to demonstrate the transformation of RO frequencies to statistically normal frequencies with high reliability through the implementation of the proposed technique. Additionally, it is shown via MATLAB simulation that the technique nullifies the effects of spatial systematic variation on the average RO frequencies extracted from four different r-ROPUF structures. Finally, the response bits generated from each r-ROPUF structure successfully passed the entire National Institute of Standards and Technology (NIST) statistical tests for randomness and exhibited true randomness and higher reliability comped to earlier techniques.