Security-based malicious flow monitoring and prediction using a modern cascaded network model

Connectivity is a rudimentary element of the Internet of Things (IoT) that enables the ability to execute handshake communication among networks. Such networks are vulnerable to various aspects, including malicious node operations, excessive traffic impacts, and fraudulent communications, among others, which can degrade network performance. Among the aforementioned, primarily malicious node identification and network analysis have a major impact on extending the research possibilities. Numerous contributions by practitioners with evolving ML Technologies addressed the underlying issues, leaving few coins unturned, which pave the way for our research. Dominant parameters, such as ideal feature selection, dictate the efficiency of ML Algorithms, aiding in improved network forecasting through the classification process. This study focuses on modeling Binary-Krill Herd Optimization (B-KHOA), a feature selection methodology built on correlation mapping that chooses relevant attributes for precise malicious traffic prediction in an IoT network. To forecast network traffic, a classification process that validates characteristics is carried out using Cascaded Learning Techniques. The results of the experiment provide greater prediction accuracy than the currently used ML techniques. The performance of multiple internet traffic datasets and other learning methodologies is used to evaluate and compare the suggested model.