Using warping for privacy protection in video surveillance

The widespread use of digital video surveillance systems has also increased the concerns for violation of privacy rights. Since video surveillance systems are invasive, it is a challenge to find an acceptable balance between privacy of the public under surveillance and the functionalities of the systems. Tools for protection of visual privacy available today lack either all or some of the important properties such as security of protected visual data, reversibility (ability to undo privacy protection), simplicity, and independence from the video encoding used. In this paper, we propose an algorithm based on well-known warping techniques (common for animation and artistic purposes) to obfuscate faces in video surveillance, aiming to overcome these shortcomings. To demonstrate the feasibility of such an approach, we apply warping algorithm to faces in a standard Yale dataset and run face detection and recognition algorithms on the resulted images. Experiments demonstrate the tradeoff between warping strength and accuracy for both detection and recognition.

I. Introduction

Recent rapid adoption of digital video surveillance systems, especially in public spaces and communities, has significantly increased the concern for protection of individual privacy. Typical surveillance systems are non-discriminative, surveying everyone and everything, which poses a threat to the human rights, to privacy, and fundamental individual freedoms [1] [2]. Growing number of privacy abuses and ignorance of privacy laws increase the public tension preventing full acceptance of surveillance systems. Many privacy advocates worry that the abuses of video surveillance may outweigh its benefits. Moreover, with the latest progress in video analytics (detection, recognition, and tracking), in combination with personal information from web and social networks (which is more and more easily available), the emerging multi-modal surveillance systems pose a serious threat to fundamental rights to privacy. Therefore, there is a strong demand in user-centric solutions [3], which specifically focus on protection of privacy in video surveillance systems.