Visual Privacy

With the growth and accessibility of mobile devices and internet, the ease of posting and sharing content on social media networks (SMNs) has increased exponentially. Many users post images that contain “privacy leaks” in regard to themselves or someone else. Privacy leaks include any instance in which a transfer of personal identifying visual content is shared on SMNs. Private visual content (images and videos) exposes intimate information that can be detrimental to your finances, personal life, and reputation. Private visual content can include baby faces, credit cards, social security cards, house keys and others. We find that monitoring techniques are essential for the improvement of private life and the development of future techniques. According to Pew Research Center, 79 percent of online Americans use Facebook, 32 percent of online Americans use Instagram, and 24 percent of online Americans use Twitter [1]. Any content posted to social media networks (SMNs) can be lost to someone else even after removal of the content. Stolen visual content can then be used as a transport vector for other types of cyber-attacks or social engineering [2].

We investigate (1) how pervasive social media-based privacy visual content leaks are and (2) what reasonable mitigation strategies can be developed to detect and minimize these leaks. We use deep learning techniques to identify “private” information. To gauge a typical user’s privacy leaks, we will incorporate privacy scoring. A privacy score computes an individual user’s probable exposure regarding their visual content leaks [3, 4]. On SMNs, content can be without any pre-screening or post-screening procedures, with this endeavor we plan to implement screening mitigation techniques that will secure users’ information.

  1. Client side. Users will download a third-party application to be installed with various SMN applications on electronics to prevent posting of potential leaks. This third-party application will pre-screen content before it can be posted on SMNs.
  2. Privacy Patrol. This is a crawler that will randomly look at users’ pages, screening for privacy leaks and alerting users of various potential leaks.
  3. Chaperone bot. Users can add a chaperone bot as a friend on SMNs. The chaperone bot will give users friendly suggestions based on type and frequency of privacy leaks on SMNs.
  4. Category Tag. Users select the category that the content belongs to before being uploaded to SMNs. Once tagged, an automated system will check for content compliance with tag. If it does not fit the category, the user is notified of what category tag the image should have.
  5. Privacy Score. Users will be monitored based on privacy score given by bots. The bot will check the users’ content after posting to remove any leaks.
  6. Server side. The SMN will screen visual content before uploading to platform. We suggest collaboration with SMNs to provide enforcement of user compliance and techniques.
  7. Interception. With the SMN applications, users will agree to let the SMN intercept the camera and gallery to flag and block content that should not be selected for posting.


  1. S. Greenwood, A. Perrin, and M. Duggan, “Social media update 2016,” Pew Research Center 11 (2016).
  2. H. Wilcox and M. Bhattacharya, “A framework to mitigate social engineering through social media within the enterprise,” in Industrial Electronics and Applications (ICIEA), 2016 IEEE 11th Conference on, pp. 1039–1044, IEEE.
  3. T. W. Grandison, S. Guo, K. Liu, M. Maxmilien, D. L. Richardson, and T. Sun, “Providing and managing privacy scores,” July 11, 2017. US Patent 9,704,203.
  4. K. Liu and E. Terzi, “A framework for computing the privacy scores of users in online social networks,” ACM Transactions on Knowledge Discovery from Data (TKDD) 5 (2010) no. 1, 6.