Influence of noise-based perturbation on recommendation application

A typical service provided by a Smart City is a Home Energy Management System (HEMS), an automated system that uses sensor networks to collect usage of home appliances and devices such that the collected data are accessible by the local government. For example, Smart meters, installed within individual homes, can send measures of electricity consumption for analysis in the context of the global community. Such data can be used by a forecasting tool to avoid electrical outage and improve the overall management of electricity usage by the community. However, the data collected may be protected by individual privacy laws and it is possible that the network may be violated by intruders to gain access to the lifestyle of individuals within their home. For example, an intruder may be able to identify active appliances and the number of individuals present within the residence. There is a requirement to protect the data emanating from these sensors to preserve this individual privacy information. Thus, a data concierge service capable of analyzing individual HEMS data is required; such a service would improve the lifestyle by allowing access to HEMS data by a third-party analyst. Although such a service may not require raw data, one must evaluate the trade-off between the preservation of privacy information and the effectiveness of the impact of the data services on the entire community. In this paper, we implemented a prototype application in conjunction with two noise-based perturbations. The application recommends improvements to household electricity usage based on its measured electricity consumption collected by smart meters. The perturbations provide a method to support Privacy Preserving Data Mining (PPDM) techniques. Evaluation of the performance results of the prototype applications illustrates the influence of the PPDM in improving the accuracy of said application.