INTRUSION DETECTION IN SMART ENERGY METER

Abstract

Smart metres and the Internet of Things (IoT) were increasingly used to replace conventional analogue metres in today’s modern smart home. It converts collected data of meter readings into digital format. The data can be delivered wirelessly, which decreases the amount of human labour required. Smart Meters, on the other hand, bring a slew of new ways to steal electricity. Using advanced tools or cyberattack techniques, malicious users can break into smart meters. Every year, this illegal conduct results in a significant financial loss. Energy theft detection techniques face a difficult task as a result of this. The advance metering infrastructure (AMI) has the capability of monitoring each consumer’s consumption details, tracking their patterns of consumption, billing them, and detecting variationsWith the help of the smart grid’s communi cation capabilities, utilities have been able to save their customers’ usage details. This database can be used to develop a theft detection model. Artificial intelligence based technologies are widely used in AMI, which deploys machine learning algorithm to detect prospective electricity thieves, frequentlyThe most common classification approaches involve utilising labels to iden tify unusual trends in customers’ previous electricity usage data and then detecting possible electricity theft behaviors. In this work,the supervised learning techniques were used to detect electricity theft. To assess classification accuracy, comparison of several machine learning clas sifiers such as Support Vector Machine, Nave Bayes, Decision Tree, and RandomForest,is also presented and the classification is accurately done by Deision Tree Classifier(99.67 Percent age).Unsupervised learning techniques such as Kmeans and DBSCAN are also used to quickly spot irregularities in the readings.The effectiveness of the models were examined using the data from simulated attacking system