Comparative Evaluations of Privacy on Digital Images

Abstract

Privacy preservation on social media is a societal issue nowadays. In recent years, with the continuous occurrence of the privacy leaks of user information and file, privacy and security issues have received unprecedented attention. Albeit a slew of mechanisms one available in protecting sensitive individual data, there are inadequate solutions to the critical concerns on privacy violations. Furthermore, the approaches of evaluating the potential privacy risks on social networking activities have not been yet paid enough attention. In order to preserve privacy effectively, the content is released safely on social media. This thesis introduces the necessity of protecting the image privacy and effective protection methods. The problems to be investigated that need to be solved urgently are put forward. The key factors affecting privacy are probed in depth. Also, the computer vision technology plays an essential role in the image privacy. Moreover, the theory of differential privacy is adopted, which can protect the image analysis data for broader research and cooperation. We combine qualitative method with AHP (Analytic Hierarchy Process) model to provide a more reasonable measure of privacy weights. Resultant analysis of the survey also provides metrics for evaluating privacy accuracy. The experimental results demonstrate that the model of image privacy evaluation proposed in this thesis can effectively and accurately measure the level of image privacy. Thus, the degree of picture privacy can be intuitively measured and the privacy can be adequately protected.