DMPF-CADA: A Confidence-Driven Adaptive Device Authentication With Dual-Modal Physical Fingerprints in Cloud–Edge IIoT

Secure device authentication is crucial in Cloud-Edge Collaborative (CEC) Industrial Internet of Things (IIoT) environments, where numerous electromechanical devices (e.g., industrial robots, motorized spindles) operate under physically vulnerable conditions. However, existing schemes struggle to balance security, decentralization, and efficiency. To address these limitations, this paper proposes DMPF-CADA, a blockchain-based multi-factor authentication framework that combines a Dual-Modal Physical Fingerprint (DMPF) and a Confidence-Driven Adaptive Device Authentication (CADA) strategy. DMPF leverages inherent device characteristics by fusing vibration spectra with Inertial Measurement Unit (IMU)-derived motion patterns, thereby enhancing robustness against physical cloning attacks. CADA dynamically adjusts authentication intensity by switching between lightweight and strong modalities according to real-time risk assessment, achieving an optimal trade-off between security and resource consumption on edge devices. The framework is implemented on Hyperledger Fabric to ensure decentralized and tamper-proof credential management. Experimental evaluations on public industrial datasets demonstrate that the proposed scheme effectively mitigates physical cloning, replay, and man-in-the-middle attacks, achieving a True Acceptance Rate (TAR) of 99.63%, a False Acceptance Rate (FAR) of 0.11%, and an average end-to-end authentication latency of approximately 188 ms. Compared with existing methods, DMPF-CADA provides stronger security guarantees and higher operational efficiency, making it highly suitable for IIoT-based CEC environments.