Project

Vehicle Location Privacy Protection in Location-based Services

Location obfuscation, which allows mobile users to report obfuscated locations instead of the exact locations, has been a dominating location privacy protection paradigm in many location-based services. Yet, the current obfuscation designs fail to address the vulnerabilities of vehicles, of which the mobility is restricted by the underlying road networks and traffic conditions. Particularly, individual vehicles’ mobility is restricted by various road environments, including local road network topology, speed limits, and traffic conditions. Built upon this insight, in this project, our objective is to 1) demonstrate that, with the road network environment and historical traffic flow information provided, vehicles’ trajectory can be inferred with high accuracy even their reported locations have been obfuscated, 2) propose a new location obfuscation paradigm to protect vehicles’ location privacy considering the vehicles’ mobility features over roads, and 3) develop, deploy, test, and refine the geo-obfuscation methods in dynamic location-based applications.

Publications: TMC 2020, CIKM 2020, IPSN 2020, ICDCS 2019

The IPSN paper has been reported by UVAToday.

This project has been reported by Rowan Today.

The MATLAB code of vehicle location tracking has been released here.

Data poisoning detection and control in crowdsourcing

Crowdsourcing is a sourcing model that leverages the collective intelligence of a large group of individuals from the Internet to solve problems. The success of many Crowdsourcing platforms is highly due to large number of available workers in the systems, which nevertheless brings several fundamental challenges: individual contributors often work remotely; maintain some degree of anonymity; are not thoroughly supervised, and work for relatively short-term rewards. Therefore, malicious contributors driven by a hidden agenda may purposely jeopardize the quality of crowdsourcing services by providing incorrect information, namely data poisoning attack. As such, thoughtful design of quality control in crowdsourcing is critically important.

Publications: CIKM 2019, AAMAS 2019, AAMAS 2018, ICDCS 2017, CIKM 2016