Blowfish Privacy: Tuning Privacy-Utility Trade-offs using Policies

Collaborators: Ashwin Machanavajjhala and Bolin Ding · This research aims to provide ways for trading-off the privacy of individuals in a statistical database for the utility of downstream analysis of data. We present a novel class of privacy definitions - Blowfish privacy, which offers a rich interface to specify this trade-off. We show significant utility improvements for k-means clustering, releasing cumulative histograms and answering range queries with mechanisms under Blowfish privacy in comparison to their differentially private counterparts. (more details)

Differentially Private Trajectory Synthesis

Collaborators: Ashwin Machanavajjhala, Divesh Srivastava, Cecilia M. Procopiuc, Graham Cormode · This research considers methods to publish locations traces with differential privacy guarantee. Our approach involves modeling the location traces as top-k mobility patterns for a bounded sensitivity, and sampling location traces from a noisy semi-Markov model learned on the top-k patterns. We present pruning strategies and efficient algorithms to post-process the noisy model which has an unbalanced hierarchical structure and to synthesize locations traces closer to the real datasets.(more details)

Coloring Privacy

This project has studied privacy in mobile-social system from another angle, visualization. Visualization creates images, graphics or animations to enhance people's understanding of information....(more details)

GPU Computing

Driven by the high computation power and affordability of Graphics Processing Units, known as GPU, we would like to explore the new role of GPU as a more general purpose processor for scientific and engineering applications.... (more details)