Research On Privacy Protection Of Electricity Consumption Data Classification And Detection Under Fog Computing

With the explosive growth of massive data in the smarl grid,the traditional centralized data processing method can no longer to the current that the data increase.Distributed technology fog computing can share the huge computing service pressure.In the processing of electricity consumption data by fog nodes,the classification of electricity consumption data is used to mine the electricity consumption rules of users in the area,which is convenient for the operator working in the smart grid scheduling.The abnormal data detection of electricity consumption data can ensure the correctness of electricity consumption data and protect the benefits of the smart grid.Therefore,classification and anomaly detection are two important uses of fog computing for electricity consumption data processing,and in the process of classification and anomaly detection processing,the semi-trusted fog nodes may easily lead to data leakage.As a widely used encryption technology,fully homomorphic can effectively protect data privacy.This paper studies the application of fully homomorphic technology in the privacy protection of electricity data classification and detection under fog computing.The details are as follows:Aiming at the problem of privacy leakage in CNN model classification of electricity consumption data under fog computing,this paper proposes a privacy protection method for fully homomorphic electricity consumption data CNN classification based on fog computing.This method proposes a three-layer framework,which is the cloud server,the fog node and the smart meters.The cloud server trains the CNN classification model in the real number domain based on the historical data marked by the proposed data labeling algorithm,obtains the optimal model parameters,and encrypts them and sends them to the fog node;the fog node obtains encrypted ciphertext electricity consumption data encrypted by the smart meter according to the input data size of the CNN classification model based on CKKS scheme,and then evaluates the calculation rules of the CNN classification model based on the CKKS scheme,and finally obtains the ciphertext classification result and uploads it to the cloud center.Based on the actual electricity consumption data in Ireland,the experiment verifies the classification performance of the proposed CNN classification model and the model classification performance in the ciphertext domain.The results show that the proposed method can ensure better electricity consumption data classification performance and better electricity consumption to protect data privacy.Aiming at the problem of privacy leakage and internal personnel attacks in the smart grid in the process of hybrid neural network electricity consumption data anomaly detection under fog computing,we proposes a privacy protection method for hybrid neural network anomaly detection under fog computing.In this method,the electricity consumption data of all smart meters in the area at the same time of collection is packaged and encrypted into one ciphertext,which is uploaded to the fog node.The fog node evaluates the calculation rules of the hybrid neural network model based on the CKKS scheme,and obtains the detection results in the real number domain through the assistance of the concentrator.Based on the detection results,the fog node marks abnormal ciphertext electricity consumption data,aggregates normal ciphertext electricity consumption data,obtains the total electricity consumption of regional ciphertext and uploads it to the control center.The control center decrypts and obtains the total electricity consumption but does not know the specific power consumption of individual users,thereby preventing internal attacks.Finally,the proposed method is verified by experiments that the proposed method can better detect abnormal behavior of electricity consumption,it can prevent data privacy leakage,and reduce the computational cost.