Abstract: Aerial Intelligent Reflecting Surfaces (IRS) by combining the advantages of aerial platforms and IRS, have the ability to flexibly control wireless channels. In mobile edge computing (MEC) systems, Internet of Things (IoT) devices offload computational tasks via wireless communication. Due to channel attenuation caused by obstacles and the broadcast nature of wireless channels, MEC systems face the risks of limited offloading rates and information leakage. A secure task offloading scheme is proposed for MEC systems assisted by aerial IRS to address these issues. In this system, IoT devices offloaded computational tasks to the MEC server within allocated time slots, while an eavesdropper attempted to intercept the information. The aerial IRS dynamically adjusted its reflection phase shifts and position to create secure channel conditions for task offloading. A joint optimization strategy was developed to optimize IRS phase shifts, time slot allocation, computation frequency, transmit power, and IRS position under energy and computation constraints. The original non-convex problem was decomposed using the block coordinate descent method, and a combination of the semidefinite relaxation and successive convex approximation methods was applied to convert it into a solvable convex problem. Simulation results show that the proposed scheme effectively improves the secure computation capability of the system.