Support vector machine based intrusion detection system with reduced input features for advanced metering infrastructure of smart grid

Security of communication network is essential for the smooth functioning of smart grid. In this paper, an intrusion detection system is proposed for early detection of threats in advanced metering infrastructure of smart grid. The proposed intrusion detection system has a multi-support vector machine classifier with mutual information based feature selection technique to detect attacks in Neighborhood Area Network (NAN) of smart grid. Mutual information technique selects the input features of classifier by analyzing the relation between different features with attacks. The developed classifier is the integration of multiple support vector machine classifiers in which each classifier detect specific attack only. The performance of developed intrusion detection system is analyzed by training and testing the classifier with ADFA-LD dataset. The proposed classifier outperforms the other machine learning approaches like artificial neural network in the detection of attacks. Simulation results demonstrate that the proposed intrusion detection approach with mutual information is well suitable for detecting attacks accurately in smart grid.