A Two-Stage TMV SRAM PUF Preselection Method with Fewer ECC Resources

The Temporary Majority Voting (TMV) method can enhance the stability of Static Random Access Memory Physical Unclonable Function(SRAM PUF), but it still incurs a significant overhead of Error Correction Code (ECC) resources. This paper proposes a two-stage TMV preselection method, where the preselected '1' and '0' cells are divided into two parts. The response is segmented into two equal-length segments, and each segment is composed of '1' and '0' cells with different bias strengths. During the reconstruction phase, both segments of the response use the same fuzzy extractor. The experimental results demonstrate that the proposed method reduces error correction resources by 48.6% by leveraging hardware resources reuse without compromising the performance of SRAM PUF. The proposed approach holds significant practical engineering value.