An instruction level attack detection method for new power load system based on LSTM

Abstract

Abstract To realize the efficient instruction level attack detection of the new power load resource management system under the load resource coordination and control interactive service scenario, this article proposes a novel LSTM adaptive sequence classification detection method combining MTM feature modeling and GAN countermeasure sample generation. In this method, the system operation instruction model based on MTM is established, and the generalized instruction features are expressed in detail. Simultaneously, large quantities of confrontation samples are generated by the GAN Network, which is similar to the real samples, and the problem of data imbalance in model training is solved. Finally, this paper proposes an adaptive coefficient-based LSTM for system instruction feature learning and testing and realizes the detection and recognition of hidden instruction level attacks in the new power load resource management system, which is more accurate and efficient than other methods.