Random invertible matrix for privacy preserving object detection. (March 2021)

• Record Type: Journal Article
• Authors: Zhang, Xiaokun; Zhang, Hongyu; Yu, Haoyang; Jin, Xin; Li, Xiaodong; Song, Chenggen; Yu, Mingxue; Fang, Xi
• Is Part Of: Computers & electrical engineering. Volume 90(2021)
• Publication Date: 2021-03
• Subjects: Image convolution -- Privacy protection -- Random sub-image -- Random matrix -- Random vector; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2021.107001
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: With the development of big data technology, privacy protection of computer vision algorithm has become a hot research field. To protect privacy of both terminal and cloud in the object detection algorithm, Jin et al. proposed a method to represent an image with multiple random sub-images. The cloud detects objects from these random sub-images instead of the original image. However, there is a fatal flaw in this method that the cloud can recover the original image according to some rules of random sub-images. In this paper, a more secure object detection algorithm based on the random matrix was proposed. The original image belonging to terminal and parameters of detection algorithm belonging to cloud were hidden in irreversible matrices, which is the key to protect privacy of both sides. In order to further reduce the computational complexity, this algorithm was improved by replacing irreversible matrices with random vectors. Graphical abstract: Highlights: Perform object detection while protecting the privacy of both the terminal and the cloud. Combine Viola-Jones object detector with random matrix/vector description method. Achieve double-blind object detection without losing any detection accuracy.