Considering the shortcomings of existing research on the problem of user reliability in crowdsourcing systems, it was assumed that each user had different reliability for different type of tasks, and on this basis, a task allocation mechanism for crowdsourcing system was designed based on the reliability of users. Firstly, an efficient task allocation mechanism was designed by using the greedy technology to maximize the profit of task publishers, and the task allocation scheme with the maximum benefit was chosen every time. Secondly, a mechanism of user reliability updating based on historical information was designed and determined by user historical reliability and the quality of the current task, and the final payment paid to the user was linked with the reliability of the user, so as to motivate the user to finish tasks with high quality continuously. Finally, the effectiveness of the designed mechanisms was analyzed in three ways:the total profit of task publishers, the task completion rate and the user reliability. The simulation results show that compared with ProMoT (Profit Maximizing Truthful auction mechanism), the proposed method is more effective and feasible, and the rate of the total benefit of task publishers is 16% higher. At the same time, it can solve the problem of user unreliability in the existing methods, and increase the reliability of crowdsourcing systems and the total revenue of task publishers.