The right to privacy has long been regarded as one of the basic universal human rights. The combination of ubiquitous sensors, wireless connectivity, and powerful recognition algorithms makes it easier than ever to monitor every aspect of our daily lives. From the use of sophisticated video surveillance systems to the theft of biometric signals, people are increasingly wary about the privacy of their multimedia data. To mitigate public concern over privacy violation, it is imperative to make privacy protection a priority in developing the next-generation multimedia processing algorithms. Due to the high dimensionality, high data-rates and stringent real-time requirements of multimedia systems, developing provably-secure privacy protection schemes for multimedia often leads to a blowup in complexity and remains impractical for most applications.
This research breaks this "efficiency barrier" of the classical cryptographical approach by investigating a new computational framework to combine distributed multimedia processing and homomorphic encryption. Building on recent results of the PI, this work develops efficient encrypted-domain processing through optimal computational procedures, parallel computation in cipher-text, small-field manipulation and encrypted data compression. In addition, by exploiting the interplay between privacy and the perceptual nature of multimedia, this work develops a provably-secure tradeoff scheme between privacy and complexity to significantly reduce the complexity and bandwidth requirements of encrypted-domain processing. The PI demonstrates the usability of this new framework through novel applications in biometric matching, object detection, speech analysis and computational photography. This work also incorporates outreach programs for high school students in rural areas, summer undergraduate research experiences, interdisciplinary postgraduate education and community outreach via television documentaries on research discovery.
The primary goal of this project is to develop novel theory and technology to protect privacy in the next-generation high-dimensional and high data-rate distributed multimedia processing. The research and educational activities sponsored by this award have the following intellectual merit and broader impact. For intellectual merit, the first accomplishment is to successfully apply key cryptographic primitives from secure multiparty computation in protecting the privacy of sensitive data used in a large number of basic multimedia protocols. These protocols include filtering, distance computation, wavelet transform, dimension reduction, similarity search and many others. Second, to break the efficiency barrier of these cryptographic primitives, novel schemes that can take advantage of various properties of the multimedia signals have been introduced, resulting in a ten-fold or higher reduction in complexity. Such schemes include k-Anonymous Quantization, which provides an optimal tradeoff between complexity and privacy, as well as anonymous Iriscode matching, which significantly reduces the size of the garbled circuit for computing hamming distance by the use of aggregate masks. Third, our research has also led to advancement in cryptography. Specifically, new game-theoretical approaches have been introduced to tackle collusion attacks often encountered in secure multiparty computation. Fourth, a number of novel research prototypes that combine both theoretical and practical aspects of the research have been developed. The first prototype is an anonymous access control system based on Iriscode, which supports user authentication via biometrics without any knowledge of the identity of the user. The second one is a privacy-aware video surveillance system in which each individual can fully control the access of his/her visual information. Fifth, the development of prototypes has resulted in research accomplishments beyond privacy technologies into computer vision and image processing. They include efficient solutions to optimal camera placement problem, robust visual and thermal foreground object detection under challenging environments, and new camera network calibration schemes. The broader impact of the project focus on the following areas. First, the award has fully or partially supported six doctoral, three master students and two undergraduate students. Thus far, two doctoral and three master students have successfully defended their theses and dissertations. Second, a total of seven journal papers (5 published and 2 submitted), two book chapters, and 11 conference papers on various topics from this research project have appeared in peer-reviewed publications. To further promote the importance of privacy research and build critical mass in the research community, the PI have also organized special issues on privacy technology in both IEEE Transactions on Information Forensics and Security as well as EURASIP Journal of Information Security. Third, privacy is a subject that straddles many different disciplines. Throughout the course of this project, new interdisciplinary collaborations have been established with researchers from political science, medicine, and cryptography both nationally (Kentucky and Pennsylvania) and internationally (Italy and China). These collaborations have resulted in joint publications and new project ideas. Fourth, beyond graduate student training, other educational and outreach efforts include a new undergraduate course on cybersecurity, technical demonstration to K-12 school children in Appalachian regions, the production of a web video documentary on privacy technology for the research channel, and the annual demonstration of the technology to the community during the University’s Engineering Open House. Fifth, technology developed in this project is being translated to medical and educational applications. Based on our work in privacy-aware surveillance, a video behavior observation system is currently being developed to observe clinically-significant behavior of children for diagnosis and treatment tracking of various developmental disorders such as autism and ADHD. The system is designed to be used in naturalistic environments including schools and clinics where privacy is a significant concern. This new effort is supported by a NSF I-Corps grant and further commercialization activities are underway.
