An Enhanced WOA and MI-Based Feature Selection Method for Attack Detection in Smart Meter Communication

Smart meter is a customer locality component of modern electrical grid. It is an easy target for the cyber-attacks. Learning algorithm-based intrusion detection system has good performance in various applications. It is a notable mechanism that provides security by early detection of intrusions from the meter traffic communication. The noisy data collected from the sources diminish the functioning of the attack finding algorithms. Thus, the feature selection algorithms are used to enhance the detection ratio of machine learning classifier. In this work, a collaborative method using wrapper-based whale optimization algorithm and filter-based mutual information is used to recognize the informational features. The identified features are feed as input to support vector machine classifier. The standard dataset ADFA-LD is employed to inspect the effectiveness of recommended method. The outputs certified the suitability of selected hybrid method for providing security to smart meter communication network.