Collusion-resistant Worker Selection in Social Crowdsensing Systems

The main idea behind social crowdsensing is toleverage social friends as crowdworkers to participate incrowdsensing tasks. A main challenge, however, is theidentification and recruitment of well-suited workers. Thisbecomes especially more challenging for large-scale onlinesocial networks with potential sparseness of the friendshipnetwork which may result in recruiting participants who arenot in direct friendship relations with the requester. Suchrecruitment may increase the possibility of collusion amongparticipants, thus threatening the application security andaffecting data quality. In this paper, we propose a collusionresistantworker selection method which aims to prevent theselection of colluders as suitable participants. For eachparticipant who is considered to be selected as suitable, theproposed method is aimed to prevent any possiblecollusion. To do so, it determines whether the selection of anew participant may result in the formation of a colludinggroup among the selected participants. This has beenachieved through leveraging the Frequent Itemset Miningtechnique and defining a set of collusion behavioralindicators. Simulation results demonstrate the efficacy ofour proposed collusion prevention method in terms ofselecting efficient collusion indicators and detecting thecolluding groups.

Keywords:

worker selection , collusion , data quality