Intrusion Detection in Power Information Network Based on Deep Learning and Cloud Computing

Abstract

With the electric power information network system has been popularized to all aspects of life, and its accompanying problems also come one after another, that is, its internal and external invasion ways are diversified, then the information security problems will affect the national econo-my, infrastructure and many other livelihood issues. On the one hand, the information security of power network needs the general characteristics of computer information security, while considering the characteristics of high security. According to the structure of power information network, this paper proposes an intrusion detection model to detect power information network, which uses both deep learning theory and cloud computing platform. This model not only uses parallel anomaly de-tection and misuse detection methods, but also can deal with the problem that a single misuse detec-tion model can't detect new attack variants. At the same time, it can capture a large number of data flows of power information network, learn and extract the essential characteristics of data flows through the deep feature learning and extraction ability of deep learning, and quickly and accurate-ly detect through the platform Invasion. Based on the analysis of all kinds of intrusion behaviors in power information network, this paper proposes a data extraction and analysis method based on Hadoop, which improves the accuracy of intrusion detection by analyzing massive captured data flow packets and generating characterization data. At the same time, considering the important factors affecting the detection of power information network intrusion feature selection, we study the auto-matic encoder algorithm using spark platform. According to the experimental results, it is found that this method can improve the feature selection of power intrusion detection system.