Secure and Scalable Face Retrieval via Cancelable Product Quantization

2025-08-31

AI Summary

To address privacy leakage risks from third-party facial image retrieval, this paper proposes an efficient revocable product quantization framework. The method integrates revocable biometric encoding with a hierarchical two-stage retrieval architecture: the first stage performs high-throughput candidate pruning using lightweight encrypted indices; the second stage conducts fine-grained similarity ranking directly in the ciphertext domain. A customized encryption mechanism and ciphertext-domain similarity computation are designed to ensure homomorphic security while substantially reducing computational overhead. Extensive experiments on multiple benchmark datasets demonstrate that the proposed approach achieves optimal trade-offs among retrieval accuracy (8.2% improvement in mAP@10), response latency (93% reduction versus HE-based baselines), and robustness against re-identification attacks. To the best of our knowledge, this is the first work enabling large-scale, real-time facial image retrieval with strong privacy guarantees.

Technology Category

• Computer Vision: Image and Video Retrieval
• Machine Learning: Privacy
• Search and Optimization: Distributed Search

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Despite the ubiquity of modern face retrieval systems, their retrieval stage is often outsourced to third-party entities, posing significant risks to user portrait privacy. Although homomorphic encryption (HE) offers strong security guarantees by enabling arithmetic computations in the cipher space, its high computational inefficiency makes it unsuitable for real-time, real-world applications. To address this issue, we propose Cancelable Product Quantization, a highly efficient framework for secure face representation retrieval. Our hierarchical two-stage framework comprises: (i) a high-throughput cancelable PQ indexing module for fast candidate filtering, and (ii) a fine-grained cipher-space retrieval module for final precise face ranking. A tailored protection mechanism is designed to secure the indexing module for cancelable biometric authentication while ensuring efficiency. Experiments on benchmark datasets demonstrate that our method achieves an decent balance between effectiveness, efficiency and security.

Problem

• Secure face retrieval risks user privacy from third-party outsourcing
• Homomorphic encryption is too slow for real-time face retrieval
• Proposing cancelable quantization for efficient encrypted face matching

Innovation

• Cancelable Product Quantization framework
• Two-stage hierarchical retrieval architecture
• Cipher-space ranking with biometric protection