Resilience Against Bad Mouthing Attacks in Mobile Crowdsensing Systems via Cyber Deception

Mobile Crowdsensing System (MCS) applications deploy rating feedback mechanisms to help quantify the trustworthiness of published events which over time improve decision accuracy and establish user reputation. In this paper, we first show that factors such as sparseness, inherent error probabilities of rating feedback labelers, and prior knowledge of the event trust scoring models, can be used by strategic adversaries to hijack the feedback labeling mechanism itself with bad mouthing attacks. Then, we propose a randomized rating sub-sampling technique inspired from moving target defense and cyber deception to mitigate the degradation in the resulting event trust scores of truthful events. We offer a game theoretic strategy under various knowledge levels of an adversary and the MCS in regards to picking an optimal sub-sample size for bad mouthing attacks and event trust calculations respectively, by using a vehicular crowdsensing as a proof-of-concept.