Hierarchical RLWE-based transformation for privacy-preserving and revocable fingerprint template generation with embedded hardware validation

In this paper, we propose a hierarchical Ring-LWE-based framework for generating cancellable fingerprints by employing the ring-LWE based transform. In the proposed method, the process starts with using the algorithm for generating quaternion feature representation for fingerprints based on intensity, orientation, frequency, and phase properties of fingerprints. The generated quaternion features are patched and quantized for creating the polynomial ring representations needed for the Ring-LWE encryption system. The two-tiered hierarchical Ring-LWE transform is used for generating privacy-preserved and revocable transformed templates intended to support cancellable biometric properties including practical unlinkability, diversity, and reconstruction resistance. Furthermore, we employ deep learning-based classification techniques to identify real fingerprint samples from imposter ones. We perform experiments on some benchmark fingerprint databases, namely FVC2002 and FVC2004. The results obtained show that the proposed system has high accuracy and low error rate. Furthermore, empirical security evaluations including cross-key unlinkability analysis, blind reconstruction attack analysis, and template collision analysis demonstrate practical resistance against template linkage and reconstruction attempts under the considered threat settings. To validate deployment feasibility, the proposed framework is additionally implemented on a Raspberry Pi platform, demonstrating low inference latency suitable for resource-constrained environments. Overall, the proposed framework provides a lightweight post-quantum-inspired cancellable biometric transformation framework that combines RLWE-inspired theoretical motivation with empirical system-level security evaluation, demonstrating reduced reconstruction and linkage capability under the evaluated threat model.