Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pura: An Efficient Privacy-Preserving Solution for Face Recognition

Guotao Xu, Bowen Zhao, Yang Xiao, Yantao Zhong, Liang Zhai, Qingqi Pei

TL;DR

Pura addresses privacy concerns in outsourced face recognition by enabling non-interactive, encrypted-distance computation on a twin-server threshold Paillier framework. It introduces secure protocols BatchSquare and SMIN families to perform multiplication, squaring, and minimum operations over ciphertexts, coupled with an offline precomputation mechanism to boost online performance. The system achieves recognition accuracy indistinguishable from plaintext baselines while significantly reducing storage and runtime compared with BFV-based approaches, including up to 16x speedups in large-scale scenarios. The approach provides strong privacy guarantees, leveraging semi-honest non-colluding servers and ISO/IEC 24745 confidentiality, and demonstrates practical viability through comprehensive experiments on public face datasets.

Abstract

Face recognition is an effective technology for identifying a target person by facial images. However, sensitive facial images raises privacy concerns. Although privacy-preserving face recognition is one of potential solutions, this solution neither fully addresses the privacy concerns nor is efficient enough. To this end, we propose an efficient privacy-preserving solution for face recognition, named Pura, which sufficiently protects facial privacy and supports face recognition over encrypted data efficiently. Specifically, we propose a privacy-preserving and non-interactive architecture for face recognition through the threshold Paillier cryptosystem. Additionally, we carefully design a suite of underlying secure computing protocols to enable efficient operations of face recognition over encrypted data directly. Furthermore, we introduce a parallel computing mechanism to enhance the performance of the proposed secure computing protocols. Privacy analysis demonstrates that Pura fully safeguards personal facial privacy. Experimental evaluations demonstrate that Pura achieves recognition speeds up to 16 times faster than the state-of-the-art.

Pura: An Efficient Privacy-Preserving Solution for Face Recognition

TL;DR

Pura addresses privacy concerns in outsourced face recognition by enabling non-interactive, encrypted-distance computation on a twin-server threshold Paillier framework. It introduces secure protocols BatchSquare and SMIN families to perform multiplication, squaring, and minimum operations over ciphertexts, coupled with an offline precomputation mechanism to boost online performance. The system achieves recognition accuracy indistinguishable from plaintext baselines while significantly reducing storage and runtime compared with BFV-based approaches, including up to 16x speedups in large-scale scenarios. The approach provides strong privacy guarantees, leveraging semi-honest non-colluding servers and ISO/IEC 24745 confidentiality, and demonstrates practical viability through comprehensive experiments on public face datasets.

Abstract

Face recognition is an effective technology for identifying a target person by facial images. However, sensitive facial images raises privacy concerns. Although privacy-preserving face recognition is one of potential solutions, this solution neither fully addresses the privacy concerns nor is efficient enough. To this end, we propose an efficient privacy-preserving solution for face recognition, named Pura, which sufficiently protects facial privacy and supports face recognition over encrypted data efficiently. Specifically, we propose a privacy-preserving and non-interactive architecture for face recognition through the threshold Paillier cryptosystem. Additionally, we carefully design a suite of underlying secure computing protocols to enable efficient operations of face recognition over encrypted data directly. Furthermore, we introduce a parallel computing mechanism to enhance the performance of the proposed secure computing protocols. Privacy analysis demonstrates that Pura fully safeguards personal facial privacy. Experimental evaluations demonstrate that Pura achieves recognition speeds up to 16 times faster than the state-of-the-art.

Paper Structure

This paper contains 28 sections, 7 theorems, 11 equations, 7 figures, 4 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Threshold Paillier Cryptosystem
  5. Face Recognition
  6. System Model and Threat Model
  7. System Model
  8. Threat Model
  9. Secure Computing Protocols
  10. BatchSquare
  11. SMIN
  12. Pura Design
  13. Problem Formulation
  14. Overview
  15. Registration Phase
  16. ...and 13 more sections

Key Result

Theorem 1

For two random numbers $x_0, x_1 \in [-2^\ell, 2^\ell]$, $x_0 + r_0$ and $x_1 + r_1$ are considered computationally indistinguishable, where $r_0, r_1 \stackrel{\$}{\gets} \{0,1\}^\sigma$. More precisely, given two random numbers $x_b \in [-2^\ell, 2^\ell]$ and $r \stackrel{\$}{\gets} \{0, 1\}^\sigm

Figures (7)

  • Figure 1: System model.
  • Figure 2: System overview of Pura.
  • Figure 3: Feasibility evaluation on LFW and WebFace.
  • Figure 4: DET curve comparison of Pura with baseline.
  • Figure 5: Comparison of storage costs of a single server among three solutions.
  • ...and 2 more figures

Theorems & Definitions (12)

  • Theorem 1
  • Corollary 1
  • Proof 1
  • Corollary 2
  • Proof 2
  • Theorem 2
  • Corollary 3
  • Proof 3
  • Corollary 4
  • Proof 4
  • ...and 2 more