AutoPrivacy: Automatic privacy protection and tagging suggestion for mobile social photo. Issue 76 (July 2018)

• Record Type: Journal Article
• Authors: Wei, Zhuo; Wu, Yongdong; Yang, Yanjiang; Yan, Zheng; Pei, Qingqi; Xie, Yajuan; Weng, Jian
• Is Part Of: Computers & security. Issue 76(2018)
• Page Start: 341
• Page End: 353
• Publication Date: 2018-07
• Subjects: Mobile social media -- Privacy protection -- Image tagging -- Multimedia encryption -- Computer vision; Computer security -- Periodicals; Electronic data processing departments -- Security measures -- Periodicals; 005.805
• DOI: 10.1016/j.cose.2017.12.002
• Languages: English
• ISSNs: 0167-4048

Abstract

Highlights: AutoPrivacy distinguishes intended and unintended human faces with their temporal and spatial characteristics. We propose a novel encryption scheme to conceal unintended human faces such that authorized users can access blur regions. AutoPrivacy utilizes recognition model to classify intended human faces based on existing tagged faces database. Implement AutoPrivacy at Android platform and verify its performance. Abstract: With the increasing computing and storage capabilities, smart mobile devices are changing our daily lives and are emerging as the dominant computing platform for end-users. It is popular among the mobile users to take photos including selfies whenever and wherever they like, and further the captured photos are shared to their friends through social networks such as Facebook and WeChat. However, an increasing issue with the large number of photos taken by a mobile user is local photo management, e.g., image searching among the photos without image tags. Another issue is that photo sharing on social networks may infringe on the privacy of the unintended human objects in the images. In this paper, an automatic privacy protection and tag suggestion system, AutoPrivacy, is proposed for mobile social images. In particular, AutoPrivacy attempts to exploit sensors signatures, image processing, and the recognition model to achieve automatic privacy protection for the unintended human objects and tagging suggestion for the intended human objects. We utilize public album data of volunteers from Facebook to test the proposed automatic system, and the experimental results on an Android platform show that AutoPrivacy can perform real time detection of intended/unintended human objects and in turn provide accurate privacy protection for unintended human objects, while the tagging suggestion for the intended human objects is efficient requiring less additional storage.