Real Time Electricity Theft Detection With Privacy

The stealing of electricity occurs frequently in smart grid,and the illegal users steal electricity to reduce the cost of electricity.The behavior of stealing electricity not only disrupts the order of residents’ use of electricity and causes economic losses to power supply enterprises,but also causes fire or electric shock and other safety hazards,threatening the lives and property of the surrounding people.Therefore,it is necessary to detect electricity theft.With the wide application of smart electricity meters,users can upload real-time electricity consumption data to the data center,which provides favorable conditions for the detection method of stealing electricity based on data analysis.The data center will release users’ electricity consumption data to the detector for data analysis,so as to identify users suspected of stealing electricity.In recent years,scholars have done a lot of research work in the field of electricity theft detection.At present,there are two types of solutions in the field of electricity theft detection: hardware-based solutions and data-driven solutions.Hardware-based solutions focus on designing specific metering equipment and infrastructure to easily detect electricity theft.Typical electricity theft detection equipment includes smart meter with heat sensor,radio frequency identification tag,etc.However,the high cost of deploying such devices,their vulnerability to severe weather conditions,and the difficulties of routine maintenance limit hardware-based electricity theft detection solutions.The data-driven scheme focuses on machine learning,through which users suspected of stealing electricity can be quickly locked,thus saving a lot of manpower for manual investigation.However,the existing detection schemes based on machine learning regard the edge nodes and the data center and the detector as trusted.Users directly publish the raw electricity consumption data to the data center by the edge nodes,and the data center directly publishes the raw electricity consumption data to the detector for detection,which will expose users’ privacy.Detectors sometimes sell users’ electricity consumption data to illegal organizations or individuals for profit.Illegal organizations or individuals can analyze users’ daily habits through these electricity consumption data,and even predict which day users are at home or not at home,thus leading to the occurrence of burglary and other crimes.Therefore,how to protect users’ data privacy while completing the detection of electricity theft is a problem worth studying.From the perspective of privacy in electricity theft detection,this paper proposed three kinds of electricity theft detection schemes to protect user data privacy,and the specific research is as follows:(1)Detection of electricity theft with fault tolerance and privacy.The traditional n-source anonymity raw data collection protocol can ensure the rawness and unlinkability of the data.The rawness of the data allows the data can be used for data analysis,so as to identify users suspected of stealing electricity.Unlinkability ensures that the detector receives the data but can not trace its source,thus ensuring the privacy of the user.However,the traditional n-source anonymity raw data collection protocol must have all users online at the same time.Once there is a user device failure and offline,the whole data collection protocol will not be able to operate,so that electricity consumption data can not be collected for electricity theft detection.This paper puts forward a solution for the possible failure of the system,so as to ensure that the system can still collect data and detect electricity theft normally in the case of user disconnection.(2)Electricity theft detection based on data aggregation.Traditional electricity theft detection is to analyze the electricity consumption behavior pattern of a single user for a long period of time,so as to identify abnormal users,that is,the detector is to detect the data of a single user.In the scheme proposed in this paper,the detector obtains the aggregate value of each day for many days of the user group and detects the aggregate value to identify the possible number of users stealing electricity in the group.Because the detector obtains aggregate values,it can not obtain the electricity consumption data of a single user,thus protecting the privacy of users.(3)Add noise to protect user privacy.The traditional electricity theft detection scheme directly releases the user’s raw data to the detector for detection without any data processing,which exposes the user’s privacy.In the scheme proposed in this paper,when the data is still in the user side,add noise to the data,and then release it to the detector for detection.The detector can not obtain the raw data of the user,thus protecting the privacy of the user.When the noise increases,the detection accuracy will be affected;when the noise decreases,it can not play a good role in protecting privacy.How to select a reasonable noise is the key point of this study.