Ensuring Honest Data Collection Against Collusive CSDF Attack With Binary-Minmaxs Clustering Analysis in Mobile Crowd Sensing

Mobile crowd sensing (MCS) is considered as a powerful paradigm which takes advantage of the pervasive sensor-embedded smartphones to collect data. However, MCS assumes all workers always are trusted, and thus offering opportunities for malicious workers to conduct the crowd sensing data falsification (CSDF) attack. To suppress such threat, recent efforts have been made to trust mechanism. Currently, some malicious workers can collude with each other to form a collusive clique, and thus not only increasing the power of CSDF attack but also avoiding the detection of trust mechanism. To ensure honest data collection in MCS, we must fight against such collusive CSDF attack. Noting that the duality of sensing data, we propose a defense scheme called BMCA from the design idea of binary-minmaxs clustering analysis to suppress collusive CSDF attack. In the BMCA scheme, the logic AND operation corresponding to the type of “1” and “0” historical sensing data is used to measure the similarity between any two workers. Based on this, we find the feature that collusive CSDF attackers usually hold high trust value and a low variance in their similarity vector. To detect collusive CSDF attackers, the min and max variance analysis is introduced to design a new binary-minmaxs clustering algorithm. Moreover, the BMCA scheme can perfect trust evaluation to prevent the trust value growth of collusive CSDF attackers. Simulation results show that the BMCA scheme can enhance the accuracy of trust evaluation, and thus successfully reducing the power of collusive CSDF attack against data collection in MCS.