Authors: Mukesh K. Saini; Pradeep K. Atrey; Sharad Mehrotra; Mohan S. Kankanhalli
Addresses: National University of Singapore, AS6, #05-19, 13 Computing Drive, 117417, Singapore ' University of Winnipeg, 515 Portage Avenue, Winnipeg, MB R3B 2E9, Canada ' Department of Computer Science, University of California, Irvine, CA, 92601, USA ' National University of Singapore, AS6, #05-06, 13 Computing Drive, 117417, Singapore
Abstract: Current surveillance systems record an enormous amount of video footage everyday. This video contains events and activities of real life which are useful in many applications. In this paper, we explore privacy preserving publication surveillance video footage, which requires robust privacy modelling and selection of appropriate data transformation function. Traditional privacy protection methods only consider implicit channels of privacy loss (such as facial information), ignoring other implicit channels. The proposed privacy model consolidates the identity leakage through both implicit and explicit channels. To choose data transformation function, we propose computational models for privacy loss and utility loss and study the tradeoff between these two. Experiments show that the hybrid data transformation method (using a combination of quantisation and blurring) provides the best tradeoff between privacy and utility. Furthermore, applying blurring first and then quantising gives better results.
Keywords: surveillance video; privacy preservation; privacy protection; data publication; security; privacy modelling; identity leakage; data transformation function; privacy loss; utility loss; quantisation; blurring.
International Journal of Trust Management in Computing and Communications, 2013 Vol.1 No.1, pp.23 - 51
Received: 02 Jun 2012
Accepted: 08 Jun 2012
Published online: 06 Mar 2013 *