Video surveillance has become prevalent both in public spaces, e.g. to prevent crimes, and in private areas, e.g. in order to assist the staff in assisted living communities. This leads to privacy concerns regarding the ability of third parties to create profiles and track individuals, possibly across several services. Usually, techniques such as pixelation and silhouettes are used to anonymize individuals. However, no formal treatment of privacy for video data has been proposed and current anonymization techniques are simply "best practice". To resolve this unsatisfactory state of affairs, we initiate a formal treatment of privacy in video data and propose a game-based notion for privacy in video data that is inspired by cryptographic security games. We show for an exemplary video privacy scheme that this scheme satisfies our notion with good parameters. In order to evaluate these parameters, we conduct a user study where the users essentially play the role of the adversary in the privacy game. Our approach can be used as a blueprint to evaluate the privacy of other video privacy schemes.
展开▼
