home · mobile · calendar · colloquia · 2002-2003 · 

Colloquium - Lookabaugh

Multimedia and Security Selective Encryption
Tom Lookabaugh
Department of Interdisciplinary Telecommunications

Most multimedia content is compressed to save on communication bandwidth or storage; although the cost of both has declined rapidly, the cost of computation has declined even faster so ubiquitous compression makes sense. Moreover, much multimedia content needs to be secured for privacy or for rights management. Increasingly, then, multimedia is subject to both compression and encryption operations.

Selective encryption leverages this now common combination; rather than encrypt all the content, we only encrypt a small but critical fraction, relying on the characteristics of compression algorithms to make the remainder (in the clear) unusable. This basic idea has occurred to a number of authors in scattered applications, including MPEG-1 video, wireless video, MP-3 audio, and G.723 audio. But the fundamental idea has much broader applicability, both in terms of the range of compression and encryption algorithms that can be exploited and in terms of functionality.

For example, selective encryption can be used to reduce computational complexity, a potentially important feature in any kind of portable multimedia application. (Indeed, this feature is likely to be important outside of multimedia as well.) It can allow for the overlaying of multiple security systems on a single multimedia stream with low overhead (a commercially important application in digital television). It can allow for efficient caching of bulk "in-the-clear" content close to users while providing a small fraction of encrypted content at decode time from more distant locations.

This presentation starts the process of building a comprehensive theoretical basis for selective encryption (and the more general variable encryption), and takes a cryptanalytic approach to validating the security implications; both of these are novel contributions to the field. The commercially important MPEG-2 algorithm is taken as a test case and selective encryption choices and their security consequences are experimentally analyzed. From this work, we validate the key insight that practical high performance selective encryption will be strongly dependent on the way in which the compressor is operated: a cooperative compressor can yield good security even with a low fraction of encryption, while an antagonistic compressor can easily undermine selective encryption unless much higher fractions of the content are encrypted (while a neutral compressor might do either).

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)