Research On Secure Transmission Method Of Smart Grid Based On Auto-encoder

With the advent of industry 4.0,large industrial infrastructures began to be equipped with the cyber-physical system(CPS).A smart grid is a typical example.In CPS-related technologies,wireless communication network has gradually replaced wired transmission and become an important part of the smart grid because of their low-cost and efficient remote transmission capacity.However,the use of communication networks brings more weaknesses to the cyber-physical system,making it more vulnerable to a variety of cyber-physical attacks.Different from traditional IT attacks,CPS attacks will not only lead to the disclosure of confidential information,threaten the security and reliability of data,but also cause direct and real damage to entities in physical space.With the frequent occurrence of CPS security events,the research on secure transmission method of smart grid is of great significance.In this paper,a data-driven secure transmission scheme is proposed.Based on the auto-encoder model of unsupervised learning,a secure transmission model of smart grid integrating encryption transmission and attack detection is designed.Starting from the traditional auto-encoder model and inspired by the denoising auto-encoder,in order to solve the challenges of background noise interference and unknown attack mode in the transmission process of information physical system,an modified denoising auto-encoder(MDAE)model is designed,and the model is used to construct the encryption and decryption module and attack detection module.In addition,in order to improve the detection ability of denial of service attack and replay attack injected in all channels,a synchronous random signal check code based on mixed Chebyshev mapping(MCM)is proposed.Firstly,the smart grid and typical attack model are constructed to obtain the sensor data under normal working state for training and the sensor data containing attack for testing;Then,only the normal data is used to train the MDAE model.The encoder and decoder in the MDAE model are used for data encryption and decryption respectively.Under the same framework,the residual generator is designed based on the same MDAE model to indicate the occurrence of the attack;At the same time,the sender and receiver synchronously generate random signals based on MCM.The random signals at the sender are transmitted together with the transmission signals,and a residual is formed at the receiver to improve the attack detection model.In this paper,the IEEE 118 bus system is used to verify the proposed secure transmission effect.At the same time,several comparative experiments on the current commonly used methods are set up.The experimental results show that the proposed method can realize the nonlinear asymmetric encryption of smart grid data with a large key space,and save the transmission cost by compressing the data.At the same time,no matter the attack mode is known or unknown,the model has accurate attack detection ability for attacks that destroy or do not destroy data integrity.