Generating High-Security Revocable Biometric Templates Based on Weighted Absolute Value Transformation With Random Bezier Matrix

With the widespread deployment of biometric authentication in Internet of Things (IoT) applications, protecting biometric templates while maintaining recognition performance and low resource consumption has become an important issue. However, existing revocable biometric template protection methods are mainly based on random projection, whose linear structure poses a reversibility risk when both transformation parameters and multiple protected templates are compromised. To address this issue, this paper proposes a revocable biometric template generation method, termed Weighted Absolute Value Transformation based on Random Bezier Matrices (WAVTRBM). The proposed method employs random Bezier matrices to preserve the discriminative information of biometric features during transformation, and introduces a random weight vector into the absolute value transformation to improve similarity preservation and matching performance. The method can generate both real-valued and binary protected templates, with the advantages of low storage overhead and fast computation, making it suitable for resource-constrained IoT devices and real-time authentication scenarios. Experiments on face, fingerprint, palmprint, and palm vein databases show that, compared with conventional random projection-based methods and existing absolute value transformation-based methods, the proposed method improves template security while maintaining good recognition performance and supporting efficient template generation and matching. Theoretical and experimental analyses further demonstrate that the proposed method satisfies the requirements of revocable biometrics and can resist various attacks.