With the high flexibility and mobility of unmanned aerial vehicles (UAVs), carrying mobile computing servers on UAV is able to provide effectively computation service for the ground users (GUs). However, since UAVs usually adopt line-of-sight transmission, malicious eavesdroppers can eavesdrop on the offloading data of the GUs. Thanks to non-orthogonal multiple access (NOMA), it allows wireless devices to provide cooperative jamming to the eavesdropper during GUs' offloading. In this paper, we propose NOMA assisted secure dual-UAV mobile computation offloading (MEC) networks. Specifically, two UAVs perform as the MEC server and the jamming UAV in turn. When one UAV provides MEC service to the GU, the other UAV sends interference signals to reduce the offloading information overheard by the eavesdropper. We then investigate the joint optimization of the GUs' computation offloading, GU-UAV's transmission and UAVs' position to maximize the GUs' benefit in a fairness manner. Due to the non-convexity of the formulated problem, we divide it into three sub-problems and propose algorithms to solve them. Numerical results are provided to validate the efficiency of our proposed algorithms.