Privacy-preserving face attribute classification via differential privacy

被引:0
|
作者
Zhang, Xiaoting [1 ]
Wang, Tao [1 ]
Ji, Junhao [1 ]
Zhang, Yushu [1 ]
Lan, Rushi [2 ]
机构
[1] Nanjing Univ Aeronaut & Astronaut, Coll Comp Sci & Technol, Nanjing 211106, Peoples R China
[2] Guilin Univ Elect Technol, Guangxi Key Lab Image & Graph Intelligent Proc, Guilin 541004, Peoples R China
关键词
Face attribute classification; Privacy protection; Differential privacy; Learnable privacy budget; PROTECTION;
D O I
10.1016/j.neucom.2025.129556
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The development of face attribute recognition technology has enhanced the intelligence capabilities in the retail industry. Merchants use the surveillance system to capture customers' face images, and analyze their basic characteristics to provide accurate product recommendations and optimize product configurations. However, these captured face images may contain sensitive visual information, especially identity-related data, which could lead to potential security and privacy risks. Current methods for face privacy protection cannot fully support privacy preserving face attributes classification. To this end, this paper proposes a privacy protection scheme that employs differential privacy in the frequency domain to mitigate risks in face attribute classification systems. Our main goal is to take the frequency domain features perturbed with differential privacy as the input of the face attribute classification model to resist privacy attacks. Specifically, the proposed scheme first transforms the original face image into the frequency domain using the discrete cosine transform (DCT) and removes the DC components that contain the visual information. Then the privacy budget allocation in the differential privacy framework is optimized based on the loss of the face attribute classification network. Finally, the corresponding differential privacy noise is added to the frequency representation. The utilization of differential privacy theoretically provides privacy guarantees. Sufficient experimental results show that the proposed scheme can well balance the privacy-utility.
引用
收藏
页数:10
相关论文
共 50 条
  • [1] Privacy-Preserving Annotation of Face Images through Attribute-Preserving Face Synthesis
    Shirai, Sola
    Whitehill, Jacob
    2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW 2019), 2019, : 21 - 29
  • [2] Efficient privacy-preserving classification construction model with differential privacy technology
    Lin Zhang
    Yan Liu
    Ruchuan Wang
    Xiong Fu
    Qiaomin Lin
    Journal of Systems Engineering and Electronics, 2017, 28 (01) : 170 - 178
  • [3] Efficient privacy-preserving classification construction model with differential privacy technology
    Zhang, Lin
    Liu, Yan
    Wang, Ruchuan
    Fu, Xiong
    Lin, Qiaomin
    JOURNAL OF SYSTEMS ENGINEERING AND ELECTRONICS, 2017, 28 (01) : 170 - 178
  • [4] Lightweight Privacy-Preserving Ensemble Classification for Face Recognition
    Ma, Zhuo
    Liu, Yang
    Liu, Ximeng
    Ma, Jianfeng
    Ren, Kui
    IEEE INTERNET OF THINGS JOURNAL, 2019, 6 (03): : 5778 - 5790
  • [5] Privacy-Preserving Face Recognition
    Erkin, Zekeriya
    Franz, Martin
    Guajardo, Jorge
    Katzenbeisser, Stefan
    Lagendijk, Inald
    Toftt, Tomas
    PRIVACY ENHANCING TECHNOLOGIES, PROCEEDINGS, 2009, 5672 : 235 - +
  • [6] Privacy-Preserving Monotonicity of Differential Privacy Mechanisms
    Liu, Hai
    Wu, Zhenqiang
    Zhou, Yihui
    Peng, Changgen
    Tian, Feng
    Lu, Laifeng
    APPLIED SCIENCES-BASEL, 2018, 8 (11):
  • [7] Privacy-Preserving Associative Classification
    Priyanka, Garach
    Darshana, Patel
    Radhika, Kotecha
    INFORMATION AND COMMUNICATION TECHNOLOGY FOR INTELLIGENT SYSTEMS (ICTIS 2017) - VOL 2, 2018, 84 : 245 - 251
  • [8] Privacy-preserving SVM classification
    Jaideep Vaidya
    Hwanjo Yu
    Xiaoqian Jiang
    Knowledge and Information Systems, 2008, 14 : 161 - 178
  • [9] Privacy-preserving SVM classification
    Vaidya, Jaideep
    Yu, Hwanjo
    Jiang, Xiaoqian
    KNOWLEDGE AND INFORMATION SYSTEMS, 2008, 14 (02) : 161 - 178
  • [10] DPVisCreator: Incorporating Pattern Constraints to Privacy-preserving Visualizations via Differential Privacy
    Zhou, Jiehui
    Wang, Xumeng
    Wong, Jason K.
    Wang, Huanliang
    Wang, Zhongwei
    Yang, Xiaoyu
    Yan, Xiaoran
    Feng, Haozhe
    Qu, Huamin
    Ying, Haochao
    Chen, Wei
    IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, 2023, 29 (01) : 809 - 819