Privacy-Preserving Detection of Power Theft in Smart Grid Change and Transmit (CAT) Advanced Metering Infrastructure

For energy management and billing purposes, advanced metering infrastructure (AMI) requires periodic transmission of consumer power consumption readings by smart meters to the electric utility (EU). An efficient way for collecting is change-and-transmit approach (CAT AMI) whereby are only transmitted if there an enough change in readings. CAT AMI, however, plagued malicious consumers who hack their illegally lower electricity bills falsifying These attacks on AMI could have bad economic consequences impair performance grid these used managing grids. Machine learning models can be detect false but this disclosing consumers’ EU evalaute model. However, jeopardizes privacy due fact that reveal sensitive information about lifestyles, e.g., presence or absence, appliances they use, etc. The problems detecting theft while protecting investigated paper. First, a dataset actual generate benign developped followed proposing new cyber-attacks tailored samples. Then, two deep-learning detectors using baseline model (CNN) CNN-GRU trained thefts AMI. To preserve privacy, paper develops enable evaluate detector encrypted data without being able learn Extensive experiments were carried out assess our proposal, results indicate proposal capable accurately identifying with acceptable overhead preserving consumers. Specifically, comparing CNN model, increases detection rate from 93.85% 97.14% HD 87.7% 94.28%, respectively.