With the maneuverability and mobility control of unmanned aerial vehicle (UAV), carrying mobile edge computing (MEC) servers on UAVs is able to effectively alleviate the explosive growth of data traffic pressure. However, UAV adopts line-of-sight (LOS) transmission which has broadcasting characteristics. Malicious eavesdroppers can easily take advantage of the characteristics to eavesdrop information during the UAV edge computing. Therefore, the security of the UAV-MEC systems is a challenging problem. This paper proposes a secure communication scheme for the Dual-UAV-MEC system. In the proposed scheme, UAV Server assists ground users in calculating the offloading tasks. In order to reduce the eavesdropping of offloading information by UAV Eavesdropper, Jammer sends interference signals on the ground. We aim to maximize the user's minimum secure calculation capacity by optimizing resources and trajectory of the UAV Server. We firstly transform the optimization problem into a tractable form through mathematical methods, and use successive convex approximation (SCA) and block coordinate descent (BCD) algorithms to solve it in an iterative manner. The final numerical results show that, compared with the benchmark schemes, the method proposed in this paper effectively increases the secure calculation capacity of the system.