Towards a Privacy Preserving Surveillance Approach for Smart Cities

Potent and continuous surveillance is an essential feature in modern smart city frameworks. However, with large scale video surveillance, there arise issues around maintaining the privacy of the people and the things they consider as personal (such as cars, pets, etc.). Furthermore, although privacy-preservation of individuals is increasingly being advocated with regulations like GDPR (in Europe), it is becoming more and more difficult to differentiate between public and private data. Therefore, technology needs to devise novel methods to incorporate privacy-ensuring mechanisms in the captured surveillance data streams. This paper presents a novel hierarchical framework directed towards ensuring privacy protection through information hiding in surveillance videos. Three layers, viz. prohibited, limited authority, and full control, have been defined based on the intended final usage of the data. These layers assign different levels of obfuscation to the elements in the video. For instance, when applied to a traffic scene, the proposed approach works by first detecting moving objects (motion detection), and then applying a blur filter on the moving objects to obfuscate them. The application of the pro- posed framework is demonstrated on videos captured in both indoor as well as outdoor settings. The proposed framework also compares favorably with existing privacy preservation approaches.