Cyber-Resilient Mobile Edge Computing: A Deep Neural Approach for Secure and Efficient Task Offloading

Abstract

Mobile edge computing (MEC) pushes cloud resources including computation and storage in vicinity of end devices. This effectively reduces communication latency. However, the minimum battery power of end devices limits the MEC performance. Integration of Wireless power transfer (WPT) with MEC enhances the performance by charging the end devices simultaneously while computation offloading. The problem lies in the variation of network state and channels while this offloading, which reduces the overall computation rate. To this end, in this paper, we aim to maximize the computation rate of whole MEC network using deep learning. We use a Deep Neural Network which learns from multiple episodes and decides whether to offload the task or not based on network state. This binary decision leads to remote execution of the task in case of optimal network state, and if not optimal, then leads to local execution. The experiments and results validate the effectiveness of our proposed framework.