The integration of power grid with digital technologies has brought significant advancements to energy management, however, it has also heightened the risk of energy theft. To tackle this challenge, this study proposes a federated energy theft detection framework for smart grids. The proposed framework employs a federated learning system composed of a federated server and several distributed power companies. In this system, each power company can contribute to the improvement of energy theft detection without compromising data privacy. The deep learning model used in this framework is designed using Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (BiGRU)-Attention.To evaluate the effectiveness of the proposed framework, a federated network communication gain metric has been proposed. The results of the experiments demonstrate that the proposed framework outperforms centralized approaches in terms of detection performance, particularly in the case of a single power company, where the accuracy of detection can increase by up to 2%. Furthermore, numerical results suggest that the proposed scheme exhibits considerable network communication gain and is suitable for the future smart cities. In conclusion, this study presents a novel federated energy theft detection framework that has the potential to improve the energy management system while preserving data privacy.

Attribute based encryption (ABE) is usually used for authorization to ensure data confidentiality because it can provide fine-grained access control. However, the data of smart grid is not completely divided, and there are many cases of collaborative access to data. The existing ABE based collaboration schemes have the defect that the ciphertext increases with the increase of attributes. Due to the limited storage and transmission capacity of smart grid equipment, the previous schemes are not suitable for smart grid. An attribute based collaborative access control scheme with constant ciphertext length for smart grid is proposed. By extending the collaborative nature to ABE with constant ciphertext length, the vacancy of collaborative ABE is made up. Performance analysis shows that this scheme is significantly better than the traditional collaborative ABE scheme in terms of storage and computing overhead. Security analysis shows that the scheme can ensure the confidentiality of data and support the revocation of collaborative at any time.

In smart grids, a major challenge is how to effectively utilize consumers’ energy consumption data while preserving security and privacy. In this article, we tackle this challenging issue and focus on energy theft detection, which is very important for smart grids. Specifically, we note that most existing energy theft detection schemes are centralized, which may be unscalable, and more importantly, may be very difficult to protect data privacy. To address this issue, we propose a novel privacy-preserving federated learning framework for energy theft detection, namely, FedDetect. In our framework, we consider a federated learning system that consists of a data center (DC), a control center (CC), and multiple detection stations. In this system, each detection station (DTS) can only observe data from local consumers, which can use a local differential privacy (LDP) scheme to process their data to preserve privacy. To facilitate the training of the model, we design a secure protocol so that detection stations can send encrypted training parameters to the CC and the DC, which then use homomorphic encryption to calculate the aggregated parameters and return updated model parameters to detection stations. In our study, we prove the security of the proposed protocol with solid security analysis. To detect energy theft, we design a deep learning model based on the state-of-the-art temporal convolutional network (TCN). Finally, we conduct extensive data-driven experiments using a real-energy consumption data set. The experimental results demonstrate that the proposed federated learning framework can achieve high accuracy of detection with a smaller computation overhead.