Privacy-Preserving Smart Metering in Smart Grids: A Systematic Survey and Multidimensional Taxonomy

Abstract

Smart meters support billing, grid operation, demand response, analytics, and distributed-energy services, but their fine-grained measurements can also reveal persistent patterns of household and business activity. Evaluating privacy-preserving solutions therefore requires more than identifying the mechanism used; it requires examining what information each observer receives, which result the application releases, and which trust and lifecycle assumptions sustain the claimed protection. This paper presents a systematic survey of privacy-preserving smart metering based on 23 prior surveys and 150 primary studies published from 2010 to the final search cutoff. Searches of IEEE Xplore, Scopus, and the Web of Science Core Collection were completed on 29 August 2026 and supplemented by citation, title, and publication-version searches, backward and forward snowballing, and targeted Google Scholar searches. We develop a multidimensional taxonomy that connects applications and authorized outputs with information flows, privacy mechanisms, adversary and collusion models, guarantees, residual leakage, operational resilience, efficiency, and evidence maturity. The synthesis shows that mechanism labels alone are poor indicators of assurance. Input confidentiality does not control inference from exact or repeated outputs; data locality does not protect model updates or released models, and anonymity may coexist with linkage through timing, location, settlement, or persistent state. Although later work supports richer functions, stronger adversaries, dynamic participation, and hybrid architectures, construction-level assurance, longitudinal output privacy, coalition and state compromise, lifecycle-safe operation, cross-layer composability, and reproducible deployment evidence remain unevenly addressed. The survey converts these findings into an evidence-derived research agenda and a requirement-driven framework for selecting protocol families according to the required output, acceptable leakage, trust boundary, lifecycle conditions, and available evidence.