Scrambling for Privacy Protection in Video Surveillance Systems

Only available on StudyMode
  • Download(s) : 59
  • Published : September 27, 2011
Open Document
Text Preview

Abstract—In this paper, we address the problem of privacy protection in video surveillance. We introduce two efficient approaches to conceal regions of interest (ROIs) based on transform-domain or codestream-domain scrambling. In the first technique, the sign of selected transform coefficients is pseudorandomly flipped during encoding. In the second method, some bits of the codestream are pseudorandomly inverted. We address more specifically the cases of MPEG-4 as it is today the prevailing standard in video surveillance equipment. Simulations show that both techniques successfully hide private data in ROIs while the scene remains comprehensible. Additionally, the amount of noise introduced by the scrambling process can be adjusted. Finally, the impact on coding efficiency performance is small, and the required computational complexity is negligible. Index Terms—Privacy, selective encryption, surveillance, video processing.



IDEO surveillance systems are omnipresent nowadays, with large systems in use in strategic places such as public transportation, airports, city centers, or residential areas. The prevailing sense of insecurity at the beginning of this century, with terrorist threats and high criminality, renders the intensive use of video surveillance tolerable despite its Orwellian big brother nature. However, people have a legitimate fear of this invasion of their personal privacy, with this objection slowing down a wider acceptance of video surveillance systems. In this paper, we address the issue of privacy protection in video surveillance, with a goal to be able to conciliate the needs of video surveillance with the objection of privacy invasion. This issue has been previously addressed in [1]–[10]. In [1], the scene is represented using an object-based representation. Depending on the end-user access control authorizations, the system subsequently renders a modified version of the video where some objects are masked out. Hence, privacy-sensitive data is not transmitted. In [2], privacy filters, expressed using a privacy grammar, are introduced. These filters are applied on incoming video sensor data, preventing access to privacy-sensitive information. In [3], it is postulated that face recognition techniques pose the threat to automatically identify people in a video surveillance scene, hence increasing the invasion of privacy. This issue is addressed by introducing an algorithm to de-identify faces such that many facial characteristics are preserved but the face cannot be reliably recognized.

Manuscript received November 14, 2007; revised March 8, 2008. First published July 9, 2008; current version published August 29, 2008. This paper was recommended by Guest Editor M.-T. Sun. The authors are with Emitall Surveillance SA, CH-1820 Montreux, Switzerland (e-mail:; Color versions of one or more of the figures in this paper are available online at Digital Object Identifier 10.1109/TCSVT.2008.928225

In [4], encryption is used to conceal faces. The process is invertible only for authorized users in possession of the secret encryption keys, hence preserving the privacy of people under surveillance. The problem of privacy for JPEG 2000 video [11] is tackled in [5] and [6]. More specifically, video analysis is applied to identify regions of interest (ROIs) corresponding for instance to people or faces. In [5], a wavelet-domain or codestream-domain conditional access control technique is proposed to subsequently scramble code-blocks corresponding to these ROIs. Alternatively, in [6], these same code-blocks are downshifted to the lowest quality layer of the codestream. By restricting the transmission bandwidth, the ROIs are decoded to a lower quality. In [7], a region-based transform-domain scrambling technique is proposed to preserve privacy. ROIs are first estimated and the corresponding transform ac...
tracking img