An Improved Differentially Private DBScan Clustering Algorithm for Vehicular Crowdsensing

Abstract

In response to the risks of privacy leakage caused by uploading vehicle participants' sensing data (such as location data or task sensing data) in the vehicle crowdsensing environment, an improved DBSCAN Clustering Algorithm (I-DP-DBScan) is proposed in the paper. The proposed algorithm is improved by optimizing the parameters Eps and MinPts, and differential privacy is adopted for guaranteeing participants' privacy. Laplace noise is added to the Euclidean distance of sensing data. Experimental results show that the I-DP-DBScan algorithm not only protects data privacy but also has good clustering efficiency.