A Novel Deep Learning Model for Privacy based Blockchain with Electricity Theft Detection in Smart Grid

Abstract

Abstract A lot of researches are being done in energy management system pertaining to intrusion detection to protect data privacy and a blockchain-based energy framework for smart-grids to detect both current and future cyberattacks. In these methods, learning-based ensemble models can assist in the identification of sophisticated malicious events while still preserving data privacy. In order to provide security, this paper proposes a privacy-based blockchain with distributed decentralized intrusion identification and electricity theft identification in smart-grids. The privacy based blockchain method is developed using BLS Short signature and hash functions. The intrusion detection method is employed by a hybrid framework leveraging Siamese Bi-LSTM for semantically discriminating malicious and authentic behaviors. In order to address the issue of class imbalance, we have used an RNN-GAN for the identification of electricity fraud. The RNN-GAN generates fake/synthesized theft samples that are very similar to actual theft instances using both supervised and unsupervised loss functions. RNN-GAN, on the other hand, even adjusts the weights of the points that are present on the accurate side of the decision boundary and keep the model from experiencing vanishing gradient problems. Additionally, batch normalization and dropout layers are used to improve the model's generalizability and speed of convergence. Our models' performance has reached great accuracy and a low error rate. Additionally, the statistical analysis demonstrates the effectiveness of the put-forth techniques.