GeoTruCrowd: trustworthy query answering with spatial crowdsourcing

With the abundance and ubiquity of mobile devices, a new class of applications, called spatial crowdsourcing, is emerging, which enables spatial tasks (i.e., tasks related to a location) assigned to and performed by human workers. However, one of the major challenges with spatial crowdsourcing is how to verify the validity of the results provided by workers, when the workers are not trusted equally. To tackle this problem, we assume every worker has a reputation score, which states the probability that the worker performs a task correctly. Moreover, we define a confidence level for every spatial task, which states that the answer to the given spatial task is only accepted if its confidence is higher than a certain threshold. Thus, the problem we are trying to solve is to maximize the number of spatial tasks that are assigned to a set of workers while satisfying the confidence levels of those tasks. Note that a unique aspect of our problem is that the optimal assignment of tasks heavily depends on the geographical locations of workers and tasks. This means that every spatial task should be assigned to enough number of workers such that their aggregate reputation satisfies the confidence of the task. Consequently, an exhaustive approach needs to compute the aggregate reputation score (using a typical decision fusion aggregation mechanism, such as voting) for all possible subsets of the workers, which renders the problem complex (we show it is NP-hard). Subsequently, we propose a number of heuristics and utilizing real-world and synthetic data in extensive sets of experiments we show that we can achieve close to optimal performance with the cost of a greedy approach, by exploiting our problem's unique characteristics.