An Efficient Intrusion Detection System using Multi-objective Genetic Algorithm and Extreme Learning Machine for Neighborhood Area Network of Smart Grid

Abstract

Intrusion detection is a major security concern in neighborhood area network of smart grid. Intrusion detection system uses classifiers for detection purpose and it suffers from the irrelevant and noisy features of network traffic. Feature selection enhances the attack detection by selecting the most informative features of network traffic as input. In this paper, the task of identifying the relevant features is represented as multi-objective optimization problem with maximization of accuracy and minimization of number of informative features as objectives. This optimization problem is solved by applying Multi-objective genetic algorithm. In this work, the extreme learning machine is used with the optimization algorithm for the selection of diverse pareto optimal solutions. The proposed system was evaluated by the standard ADFA-LD dataset. The simulation results show that the proposed system finds multiple efficient solutions with high accuracy and less number of informative features and is suitable for the application of intrusion detection.