Overview

index_image
DiffTrajectory is a differentially private trajectory analysis technique for points-of-interest recommendation to users that aims at maximizing the accuracy of the recommendation results while protecting the privacy of the exposed trajectories with differential privacy guarantees.

DiffTrajectory can perform both:
  • Points-of-interest recommendation by analyzing the underlying trajectory training data.
  • Protect the privacy of the trajectories with rigorous differential privacy guarantees.


Motivation

threat

Trajectory-based travel recommendation:

  • Trajectories of individual mobile users can be analyzed to understand travel behavior and make personalized personal travel recommendations. The mobile users implicitly recommend their visited places to the new visitors.
  • Aggregate analysis of historical trajectory data belonging to different mobile users can provide more generalized travel recommendations.
Privacy threats：
Location information of the travel destination is often associated with a semantic meaning. The disclosure of the association between a mobile user and a location may reveal private information such as the user's health condition.



Algorithm

Phase I: User-location bipartite graph representation:

  • Raw trajectory dataset.
  • Stop points detection.
  • Stop points clustering.
  • Association construction.
  • Bipartite graph.

algorithm1

Phase II: Differentially private data mining:

  • Matrix construction.
  • Noise addition.
  • Noise suppression.
  • Hyperlink-Induced Topic Search (HITS).

algorithm2








Publications

  • Chao Li, Balaji Palanisamy and James Joshi, "Differentially Private Trajectory Analysis for Points-of-Interest Recommendation", Proc. of 6th IEEE International Congress on Big Data (BigData Congress 2017), Honolulu, USA. (Best Paper Award). [PDF]


Acknowledgement

This work was performed under a partial support by the National Science Foundation under the grant DGE-1438809.


Slides