A Novel Game Machine Learning Method for Calculating Optimal Response for Edge Server

摘要

Mobile edge computing extends traditional cloud services to the edge of the network and enables edge server to handle network requests with low latency requirements. However, the edge server is closer to the terminal device with relatively limited storage capacity and computing capacity, and is more vulnerable to the invasion of attackers. To solve this problem, we proposed a game machine learning method to determine the optimal response of edge server to attackers, so as to defend against attackers. First, we used Hidden Markov Model to fit the behavior model of the attacker; secondly, due to the payoff of edge server is closely related to the attacker's behavior model, we used the gradient ascent method to maximize the payoff of edge server; finally, the optimal response of edge server was determined. Detailed experimental results showed that the new scheme can improve the payoff of the edge server and defend against attackers.