Large-scale Intelligent Network Attack Detection based on Hierarchical Symbolic Dynamic Filtering

Abstract

Smart grid technology enhances grid security, reliability, and efficiency. In order to ensure efficient and reliable power distribution, it must address new vulnerabilities brought about by digital communication technology. In this paper, a new Energy Efficient Anomaly Detection (EEAD) technique is proposed, which uses HSDF pre-processing and HMM learning. A number of subsystems are initially created within the system. Hierarchical symbolic dynamic filtering (HSDF) converts time series data into symbol sequences and then learns the causal relationship between the nominal characteristics of subsystems. Then the converted sequences will be fed to the Hidden Markov model (HMM) which detects the anomaly by calculating the occurrence probability of the current observation based on the trained network. Simulation results on an IEEE 118 bus system to verify the performance of the suggested method under various operating conditions such as False Positive Rate, Detection rate, Accuracy, and True Positive Rate.