Crowdsourcing technology enables complex tasks to be solved with the aid of a group of workers in the Internet of Things (IoT). On the one hand, crucial sensing data can be collected and processed to enhance smart IoT applications. On the other hand, crowdsourcing IoT (Crowd-IoT) is still facing threats due to the diverse quality of crowdsourced data, and especially the misbehavior of malicious workers. In this paper, we propose a Stochastic Bayesian Game (SBG) to address the Byzantine Altruistic Rational (BAR) based misbehavior, where workers' behavioral types can be deduced reasonably and the requestor can perform optimal actions accordingly by taking the long-term gain into consideration. To validate and evaluate the performance of the proposed model, we simulate various scenarios and conduct a comparison with other solutions. The numerical results show the effectiveness and feasibility of our proposed solution.