A SRAM-PUF-Based ECDSA Authentication Framework for Secure Embedded and IoT Systems

Abstract

This paper proposes a real-time secure authentication framework based on an SRAM-based Physical Unclonable Function (SRAM-PUF) integrated with the Elliptic Curve Digital Signature Algorithm (ECDSA). Instead of storing secret keys, the proposed approach dynamically regenerates device-unique cryptographic keys from the intrinsic startup behavior of SRAM, providing a hardware-rooted security mechanism. To overcome noise and instability in SRAM-PUF responses caused by temperature and voltage variations, a fuzzy extractor with adaptive error correction is employed, enabling reliable key reconstruction under changing environmental conditions. Furthermore, a helper data integrity protection scheme is incorporated to prevent manipulation and reconstruction attacks during the key recovery process. The regenerated PUF-derived key is utilized for ECDSA-based authentication, ensuring strong resistance against device impersonation, key leakage, and physical attacks. In addition, the framework supports real-time authentication by periodically verifying key consistency during system operation, enhancing robustness against fault injection and timing-based attacks. The proposed system is implemented and validated on an embedded microcontroller platform, demonstrating a lightweight, low-overhead, and practical security solution suitable for resource-constrained embedded and Internet of Things (IoT) applications.