Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Blind-Match: Efficient Homomorphic Encryption-Based 1:N Matching for Privacy-Preserving Biometric Identification

Hyunmin Choi, Jiwon Kim, Chiyoung Song, Simon S. Woo, Hyoungshick Kim

TL;DR

The paper tackles privacy-preserving biometric identification in a 1:N setting by introducing Blind-Match, an HE-based system that optimizes cosine similarity through partitioned feature vectors. It combines a two-stage workflow (enrollment and recognition) with input ciphertext expansion, HE-$C^3$, and a compression scheme, guided by a cluster architecture to enable parallel processing. Empirical results on face and fingerprint benchmarks show state-of-the-art or near-state-of-the-art accuracy (e.g., $99.63\%$ Rank-1 on LFW with $128$-D; $99.68\%$ on PolyU with $128$-D) and practical throughput (6,144 samples in $0.74$ s), validating real-time HE-based 1:N matching. The work also provides a formal analysis of the optimal partitioning parameter $N_{in}=4$ for $m=128$ and discusses deployment plans and open-source availability, underscoring its potential for privacy-preserving biometric systems in cloud environments.

Abstract

We present Blind-Match, a novel biometric identification system that leverages homomorphic encryption (HE) for efficient and privacy-preserving 1:N matching. Blind-Match introduces a HE-optimized cosine similarity computation method, where the key idea is to divide the feature vector into smaller parts for processing rather than computing the entire vector at once. By optimizing the number of these parts, Blind-Match minimizes execution time while ensuring data privacy through HE. Blind-Match achieves superior performance compared to state-of-the-art methods across various biometric datasets. On the LFW face dataset, Blind-Match attains a 99.63% Rank-1 accuracy with a 128-dimensional feature vector, demonstrating its robustness in face recognition tasks. For fingerprint identification, Blind-Match achieves a remarkable 99.55% Rank-1 accuracy on the PolyU dataset, even with a compact 16-dimensional feature vector, significantly outperforming the state-of-the-art method, Blind-Touch, which achieves only 59.17%. Furthermore, Blind-Match showcases practical efficiency in large-scale biometric identification scenarios, such as Naver Cloud's FaceSign, by processing 6,144 biometric samples in 0.74 seconds using a 128-dimensional feature vector.

Blind-Match: Efficient Homomorphic Encryption-Based 1:N Matching for Privacy-Preserving Biometric Identification

TL;DR

The paper tackles privacy-preserving biometric identification in a 1:N setting by introducing Blind-Match, an HE-based system that optimizes cosine similarity through partitioned feature vectors. It combines a two-stage workflow (enrollment and recognition) with input ciphertext expansion, HE-, and a compression scheme, guided by a cluster architecture to enable parallel processing. Empirical results on face and fingerprint benchmarks show state-of-the-art or near-state-of-the-art accuracy (e.g., Rank-1 on LFW with -D; on PolyU with -D) and practical throughput (6,144 samples in s), validating real-time HE-based 1:N matching. The work also provides a formal analysis of the optimal partitioning parameter for and discusses deployment plans and open-source availability, underscoring its potential for privacy-preserving biometric systems in cloud environments.

Abstract

We present Blind-Match, a novel biometric identification system that leverages homomorphic encryption (HE) for efficient and privacy-preserving 1:N matching. Blind-Match introduces a HE-optimized cosine similarity computation method, where the key idea is to divide the feature vector into smaller parts for processing rather than computing the entire vector at once. By optimizing the number of these parts, Blind-Match minimizes execution time while ensuring data privacy through HE. Blind-Match achieves superior performance compared to state-of-the-art methods across various biometric datasets. On the LFW face dataset, Blind-Match attains a 99.63% Rank-1 accuracy with a 128-dimensional feature vector, demonstrating its robustness in face recognition tasks. For fingerprint identification, Blind-Match achieves a remarkable 99.55% Rank-1 accuracy on the PolyU dataset, even with a compact 16-dimensional feature vector, significantly outperforming the state-of-the-art method, Blind-Touch, which achieves only 59.17%. Furthermore, Blind-Match showcases practical efficiency in large-scale biometric identification scenarios, such as Naver Cloud's FaceSign, by processing 6,144 biometric samples in 0.74 seconds using a 128-dimensional feature vector.
Paper Structure (22 sections, 3 theorems, 1 equation, 4 figures, 7 tables, 2 algorithms)

This paper contains 22 sections, 3 theorems, 1 equation, 4 figures, 7 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Biometric Recognition with HE
  4. Overview of Blind-Match
  5. Key Generation and Management
  6. Enrollment
  7. Recognition
  8. Input Ciphertext Expansion Algorithm
  9. HE-$C^3$ Algorithm
  10. Compression Method
  11. Decryption
  12. Cluster Architecture
  13. Algorithm Analysis
  14. Experiments
  15. Experimental Setup
  16. ...and 7 more sections

Key Result

Proposition 4.1

The time it takes to perform Algorithm alg:input_ctxt_expansion_methods can be calculated as $N_{in} \cdot ((T_{Mul(P),l} + T_{Res,l}) + \log(N_{in}) \cdot (T_{Rot, l-1} + T_{Add, l-1}))$.

Figures (4)

  • Figure 1: Overview of Blind-Match. Blind-Match consists of two stages: During the enrollment phase, Blind-Match divides and stores the encrypted feature vector into smaller parts. During the recognition phase, Blind-Match processes these smaller parts through multiplication and then aggregates the results using our new cosine similarity computation method.
  • Figure 2: Structure of input ciphertexts.
  • Figure 3: Enrollment of the $i$-th input ciphertext.
  • Figure 4: Input ciphertexts for HE-$C^3$.

Theorems & Definitions (3)

  • Proposition 4.1
  • Proposition 4.2
  • Proposition 4.3