Loading
Research Projects
Differentially Private Synthesis of Location Traces
|
Abstract
The study on individual mobility patterns is popular in many fields including computer science, medicine, public health, and web and telecommunication industries. However, mobility patterns learnt directly from the raw location traces could disclose sensitive information of an individual, and hence this raises privacy concerns.
In this talk, I will describe our current efforts toward synthesizing location traces that satisfy differential privacy (a state of the art privacy definition). Our approach involves sampling location traces from a "noisy" semi-Markov model learnt from the original data. These synthetic traces can be used for analysis by third party researchers instead of the real data. We encounter three important problems in model construction: utility, consistency and choice of space decomposition. I will describe these problems and my initial work towards solving them.
Advisor(s): Ashwin Machanavajjhala
Landon Cox, Jun Yang