SRAM-PUF Response Recovery and Key Reconstruction for Intelligent Metering Terminals

With the continued development of new-type power systems and intelligent metering infrastructures, field devices such as smart meters, power information acquisition terminals, and fusion terminals have become essential components for energy metering, state sensing, and data interaction. SRAM-based physical unclonable functions (SRAM-PUFs) exploit intrinsic device variations in on-chip SRAM to generate device-specific responses and provide a lightweight hardware basis for terminal identification and key generation. However, variations in operating conditions can cause SRAM response instability and reduce key reconstruction reliability. To address this issue, this paper proposes a SRAM-PUF response recovery and key reconstruction method for intelligent metering terminals. A Bidirectional Temperature-Task Continual Learning (BTCL) mechanism is developed to continuously learn response evolution patterns under different operating conditions. Local consistency constraints and SRAM-cell reliability priors are further incorporated to recover disturbed responses, while spatial interleaving and concatenated error correction are employed to improve key reconstruction stability. Experimental results show that the proposed method reduces the mean pre-ECC bit error rate (BER) to 2.14%, corresponding to reductions of 35.93% and 26.33% compared with the raw SRAM-PUF and B-Spline Continual Learning, respectively. After spatial interleaving and concatenated error correction, the mean post-ECC BER is further reduced to 8.45 × 10−4%, while the mean key error rate (KER) decreases to 2.70 × 10−2%. These results demonstrate that the proposed method effectively improves SRAM-PUF response consistency and key reconstruction reliability under varying operating conditions, providing a reliable hardware-response and key-reconstruction basis for subsequent terminal authentication applications.