Security Analysis on Locality-Sensitive Hashing-based Biometric Template Protection Schemes

Seunghun Paik (Hanyang University), Sunpill Kim (Hanyang University), Jae Hong Seo (Hanyang university)*
The 34th British Machine Vision Conference


Designing an efficient and secure biometric template protection (BTP) scheme is a long-lasting challenge, and locality-sensitive hashing (LSH) is one of the promising building blocks for designing secure BTP schemes. We find that many existing LSH-based BTP schemes are designed with an identical structure, and thus we formulate such a structure as locality-sensitive predicate to capture its key properties. This enables us to analyze the security of a wide range of LSH-based BTPs. Based on this idea, we propose a novel method that recovers feature templates from templates protected by several LSH-based BTP schemes. In particular, the recovered templates by ours have a higher purity than those recovered by the other methods in the sense that ours recovers a close template to the original template. Recovering closer templates has several advantages over the previous methods. First, we successfully cryptanalyze a recent LSH-based BTP scheme for the first time, which was not cryptanalyzed by the previous methods. Second, by combining existing face reconstruction methods, we successfully reconstruct the face image that resembles the original face image (e.g., LFW dataset). This property has not been achieved by previous attack methods. To clearly show it, we evaluate the true accept ratio (TAR) of reconstructed face images when different face images of the same identities are enrolled. Ours achieves a similar TAR (around -0.3%∼-1.4%) to the (unprotected) recognition system, but the others achieve a much lower TAR (around -84%∼-20%). To facilitate future research, our implementation code is available on github.



author    = {Seunghun Paik and Sunpill Kim and Jae Hong Seo},
title     = {Security Analysis on Locality-Sensitive Hashing-based Biometric Template Protection Schemes},
booktitle = {34th British Machine Vision Conference 2023, {BMVC} 2023, Aberdeen, UK, November 20-24, 2023},
publisher = {BMVA},
year      = {2023},
url       = {}

Copyright © 2023 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection