An Efficient Malicious User Detection Mechanism for Crowdsensing System

Abstract

Although crowdsensing has emerged as a promising data collection paradigm by applying embedded sensors in mobile devices to monitor the real world, the data quality problem is still a big issue for the existence of malicious users in the crowdsensing system. There have been many mechanisms proposed to improve the quality of submitted observations. However, they are not cost-efficient enough to be widely applied or only compatible with limited applications. Therefore, we propose an efficient malicious user detection method by developing a Hidden Markov Model, which can distinguish malicious users from normal users. We also provide a malicious user pre-detection method by using a Gradient Boosting Decision Tree model, which is targeted at the crowdsensing system in long-term operation and has new participation constantly. The experimental results based on real-world datasets reveal that the proposed method has good accuracy of the user distinguishment and shows significant changes in improving the existing truth discovery methods.