Intrusion detection in smart meters data using machine learning algorithms: A research report

The intrusion detection in network traffic for crucial smart metering applications based on radio sensor networks is becoming very important in the Smart Grid area. The network's structure for smart meters under investigation should consider important security factors. The potential of both passive and active cyber-attacks affecting the functioning of advanced metering infrastructure is studied and a novel method is proposed in this article. The proposed method for anomaly identification is efficient and rapid. In the beginning, Cook's distance was employed to recognize and eliminate outlier observations. After observations are made three statistical models Brown's, Holt's, and winter's models were used for exponential smoothing and were estimated using the provided data. Bollinger Bands with the appropriate parameters were employed to estimate potential changes in the forecasts produced by the models that were put into operation. The estimated traffic model's statistical relationships with its actual variations were then investigated to spot any unusual behaviour that would point to a cyber-attack effort. Additionally, a method for updating common models in the event of substantial fluctuations in real network traffic was suggested. The findings confirmed the effectiveness of the proposed method and the precision of the selection of the appropriate statistical model for the under-study time series. The outcomes validated the effectiveness of the proposed approach and the precision in choosing a suitable statistical model for the time series under investigation.