Activity and resolution aware privacy protection for smart water meter databases

Identifying water end-uses from household meter readings yields valuable commercial and environmental insights for both customers and water providers. However, smart water meter databases may expose sensitive information about the activities of metered households, thus raising privacy concerns that limit the possibility of making smart meter data available for research and planning. This paper considers the case where a water provider wishes to publish a database of household water meter traces to be used by water analysts, who are often external entities. This scenario presents privacy risks should an adversary gain access to this database, with the threat of uniquely identifying a household of interest and exposing its water use activities. To mitigate such risks, this paper proposes a process for activity-aware privacy protection and evaluates its effectiveness using real-world and synthetic databases. Our experimental analysis shows that water meter privacy protection is strongly dependent on the type of activity, temporal resolution, and population size. For example, in the case of a high-resolution database, we found that a large base population, small published sample, and 30-second resolution provided optimal trade-offs between privacy and information value. For a low-resolution database with a population of over 3500 households, 1-hour resolution provided strong information value and customisable privacy guarantees depending on the published sample size.