An Enhanced Genetic Algorithm-based IDS To Secure Smart Meter Communication

The earlier attack detection of Intrusion detection system (IDS) helps to neutralize the cyber-attacks and securing the sensor data communication of smart meter. Feature selection methods are employed to improve the performance of IDS. In wrapper-based feature selection methods, the informative population plays a leading role in attack detection and rapid convergence. Due to its randomly initialization nature, it is difficult to found the relevant one. Thus, the selection of good initial population is an important criterion to get the optimized result. In this work, the initial population of genetic algorithm is tuned by mutual information (MI) technique to generate the informative population for IDS. We employ a support vector machine classifier to evaluate the developed population using a dataset generated from neighborhood area network experimental testbed. The performance of the proposed method is analyzed by comparing with fully random, MI, and MI + random based population generation method. The results show that the suggested method has better detection ratio, low false-positive rate and less execution time compared to other standard methods.