Terminal Security Monitoring Based on Power Consumption Information

Abstract

The application of information technology in many fields is becoming more and more popular, but while bringing about a rapid increase in productivity, it also brings some safety issues, especially in industrial control systems. Since the industrial control system often uses a computer as the control center of some devices, once this computer is attacked, it will cause serious harm. The use of additional security software for security monitoring is not completely credible, after all, security monitoring software will also be attacked and become invalid. Therefore, the method of using some side channels and machine learning is very popular recently, especially the power consumption side channels. However, the power consumption will change with the running time of the device. If the model trained by supervised learning will fail after a few days, this paper proposes a self-learning method based on the power consumption side channel, which can be stable for a long time with a high accuracy of 97%.