Security in Smart Grid
Privacy-preserving Advanced Metering Infrastructure for Smart Grid (Lead: Siraj)
In the smart grid, consumers are concerned about their privacy that might be violated by a utility company, who has direct access to their energy consumption. The analysis of the energy consumption behavior can expose consumer sensitive information. Such information can include (but not limited to) the appliances the consumer uses or the time interval when the consumer is absent at home. This, in turn, can potentially reveal information that can be misused for a marketing purpose or burglary preparation. In order to resolve this problem, our center has been working on developing a privacy-preserving Advanced Metering Infrastructure specifically designed to protect consumer privacy. It allows energy providers to fulfill a fine-grained energy consumption analysis such as fraud detection and load monitoring. In addition, this solution incorporates time-of-use billing and satisfies computational constraints that Smart Meters are bound by. Also, the developed model of the Advanced Metering Infrastructure provides consumers with the opportunity to select the most suitable utility company based on the time-of-use prices. To sum up, the developed architecture preserves consumer privacy and, at the same time, allows to fully utilize the capabilities of Advanced Metering Infrastructure providing fine-grained energy readings.
Smart Grid Energy Fraud Detection (Lead: Siraj)
Energy fraud detection is one of the crucial parts of the Smart Grid security. In a legacy power grid, it is difficult to detect fraudulent activities (such as tapping to an electricity line or tampering with a power meter) because of lack of data to analyze. However, Smart Grid can report fine-grained energy consumption that can be utilized for an intelligent analysis by applying machine learning techniques. Machine learning have been widely used for analyzing data, learning patterns, and extracting useful information. Graduate and undergraduate students in our center have been working on employing machine learning techniques to detect anomalies corresponding to energy fraud. They have used artificial neural networks and decision trees to successfully identify simulated fraudulent activities with a high detection rate. The machine learning techniques were trained using real-world data provided by the Irish Social Science Data Archive Center. The data comprised of energy measurements and corresponding timestamps. The energy fraud they have worked with can be categorized in two groups. The first group includes fraudulent activities that simulate rogue connections (tapping to an electricity line). The second group simulates reporting less energy than actually was consumed. As a result of applying machine learning techniques mentioned above, it was demonstrated that they can detect energy fraud with a high detection rate.