8 September 2025 in IEEE Access
The proliferation of counterfeit and refurbished semiconductor devices has become a critical concern. Device identification techniques based on Physically Unclonable Functions (PUFs) have therefore attracted significant attention. Conventional PUFs exploit only die-internal process variations and thus cannot achieve board-level identification. In this work, we propose a novel method that also harnesses external process variations on the printed circuit board (PCB)—including trace geometry variations—to realize robust board-level device identification. Our method employs an on-chip capacitance sensor to measure the capacitance between signal traces and ground. A histogram of the sensor output values is generated, from which distinctive feature points in the distribution are extracted. By selecting bits that exhibit both high reproducibility and uniqueness, we construct a 30-bit device identifier. This approach eliminates the need for the large-scale measurement systems required by traditional techniques, enabling a more convenient yet high-precision identification process. Evaluation on 20 microcontroller boards demonstrates that the proposed method achieves 97.14% identification accuracy with the 30-bit identifier. The average normalized intra-device Hamming distance (Intra-HD) is 3.7 %, and the average normalized inter-device Hamming distance (Inter-HD) is 46.7 %, closely approximating ideal values. These results indicate that our technique offers a promising approach for ensuring board-level authenticity.