Toniann Pitassi
Department of Computer Science
University of Toronto

Privacy in a Big Data World: Some Recent Research

Department of Computer Science
Thursday, September 5, 2013, 1:00 pm., (note time change) Room EN-2022

---

 Abstract

In the modern world, we have tremendous opportunity to collect and analyze huge amounts of data in an unprecedented level of detail. The massive amount of data, and the results of analyzing it, can then be used to dramatically improve the quality and efficiency of life. For example, traffic patterns, collected through GPS devices, can lead to new solutions that improve the efficiency of transportation and fuel usage. With big data comes big concerns about privacy. A key question is: Can we learn from data in a way thatis beneficial to society, while still preserving our privacy?

After decades of unsuccessful efforts to characterize privacy adequately, differential privacy has emerged as a robust, mathematically rigorous characterization of privacy-preserving data release. Differentially private mechanisms allow a robust trade-off between accuracy of predictions and the privacy of individuals, using a confluence of techniques from fields such as machine learning, statistics, algorithms, and game theory. I will first survey the basics results in differential privacy, and then discuss some new research directions in this area.