Neural Edge-cloud Computing with Information Cascade Attack

In recent years, a considerable body of work has employed machine learning methods to analyze edge computing network behavior. Previous research predominantly relied on conventional machine learning or probability models to address this issue. However, these research endeavors often neglect to discuss the network’s resilience against information cascade attacks under different settings. In this paper, we extend the basic machine learning model by introducing a network with single or multiple layers of encoder and decoder, which we term the Transition Adjoint Matrix Prediction (TAMP) model. Our aim is to examine the impact of various parameters on the outcome of the random process moving towards the absorbing state. Specifically, our investigation focuses on elucidating how the network’s structure, the proportion of information cascade attacks, and network depth influence the behavior of edge computing networks. Numerical results demonstrate that the structure of the encoder and decoder renders the network more vulnerable to cascade attacks. Moreover, as the proportion of cascade attacks increases, the likelihood of reaching the absorbing state also rises. This provides a concrete strategy for analyzing this phenomenon. Notably, the results of our proposed TAMP model reveal that the depth of the edge computing network significantly affects its resistance to cascade attacks.