<p>Existing architectures for detection of electricity theft within smart grids largely transmit power consumption or generation data directly to the detection center on the users’ behalf under supervision of smart devices. Since these centers may not be trusted entirely, there exists a potential risk to users’ data privacy. In addition, smart devices may be attacked and thus may hamper the theft detection ability through tampering. Thus, this paper proposes a decentralized electricity theft and fraud detection scheme based on blockchain technologies, which does away with a third-party intervention and protecting the privacy of users. Specifically, an enhanced homomorphic encryption solution ensures consumer privacy while enabling monitoring of electricity theft and power load, alongside authentication of smart meters and other devices. An optimal detection model that integrates homomorphic encryption and Manhattan distance is proposed to enable effective theft detection. The proposed strategy achieves high accuracy in detecting theft and fraud, up to 95.4%, while also accurately estimating total regional power. Further analysis shows that the proposed scheme can effectively defend against security threats from a variety of sources and adequately protect consumers’ privacy.</p>