SBiPMU-Based Power Data Encryption Traffic Attack Forecasting

Amidst the exponential growth of information technology, the pervasive integration of Internet of Things (IoT) solutions has become an integral part of both the everyday lives and professional routines for many individuals. However, in the information systems of the power industry, the characteristics of data flow present a phased and relatively centralized situation, which makes traditional encryption behaviors and abnormal communication behavior detection technologies difficult to deal with. Although existing attack prediction methods based on neural networks have partially solved the above problems, the process entails extensive application of nonlinear transformations for data learning and profiling, which can lead to considerable attrition of feature details. Addressing this issue, this paper design a new neural network model based on bidirectional parsimonious memory unit (BIPMU), namely the bidirectional supplementary gate minimal unit (SBiPMU). In contrast to the conventional BIPMU, the SBiPMU mitigates the degradation of feature integrity stemming from nonlinear data transformations in the learning and profiling of network traffic, employing a compensatory mechanism to preserve feature richness. Then, a new attack prediction method is proposed using SBiPMU to monitor network security. Ultimately, we have constructed a pertinent prototype system and assessed its functionality. The experimental findings suggest that our proposed strategy for anticipating attacks holds a higher degree of precision in comparison to antecedent techniques.