This study explores the physical layer secrecy in an ambient backscatter communication (AmBC)-based multi-user non-orthogonal multiple access (NOMA) wireless sensor network, integrated with mobile edge computing (MEC). In this considered network, an unmanned aerial vehicle (UAV) equipped with a server hovers above an energy-constrained wireless sensor cluster, providing RF energy and assisting in task execution through the AmBC scheme. Meanwhile, a UAV eavesdropper attempts to intercept the information transmitted from these sensor nodes. To enhance the system’s secrecy, we propose a partial offloading scheme based on the non-linear harvested energy levels of the users. A closed-form expression for the secrecy successful computation probability (SSCP) is derived, considering factors such as the harvested energy for local computation, offloading and processing latency, and secrecy capacity. The impact of UAV altitude, transmit power, backscatter reflection coefficients, task division ratios, and task lengths on both security and latency is analyzed to evaluate the system’s performance. Numerical results, obtained through Monte Carlo simulations, are provided to validate the accuracy of our analysis.