IMAO-DrMACN: Threat Detection in IoT Networking for Secure Communication in Wireless Networks

Abstract

The rapid growth of the Internet of Things (IoT) has made securing wireless networks increasingly difficult. As IoT devices interconnect, the network becomes more exposed to security threats that affect confidentiality, integrity and availability. Existing threat detection systems remain limited by high false-positive rates, poor adaptability to emerging attack patterns, and degradation of performance when the statistical properties of network traffic shift over time. This research addresses the adaptability and feature-learning limitations of existing detectors by introducing a framework named Intelligent Migration Avian Optimization enabled Drift-enhanced Modified Activation-based Convolutional Neural Network (IMAO-DrMACN). The proposed IMAO algorithm performs hyperparameter tuning, so that the model can be re-tuned as the IoT environment changes; the drift mechanism allows the model to respond to significant changes in the data distribution rather than degrading silently; and the modified activation function, which introduces a learnable channel-wise slope, improves feature learning in the convolutional blocks. The framework is evaluated on the DAPT2020 and BoT-IoT datasets under a 90% training / 10% testing split. On DAPT2020 it attains an accuracy of 97.531%, a recall of 97.476%, a precision of 97.687% and an F1-score of 97.417%; on BoT-IoT it attains 96.905%, 97.763%, 95.004% and 96.364% respectively. These correspond to accuracy gains of 2.67 and 2.95 percentage points over the strongest of the eight comparison methods (AM-DrCN) on the two datasets. The results indicate that coupling adaptive hyperparameter search with a drift-aware convolutional classifier is an effective strategy for threat detection in dynamic IoT environments.