Intelligent Cyber threat detection using Optimized Deep Neural Network approach for IoT social networks

The nature of cyber threat detection in social networks of Internet of Things (IoT) is demanding in terms of high data heterogeneity, network dynamics, and changing patterns of attack. The conventional techniques of cyber threat detection usually have low accuracy, high false positive rates and do not have good scalability. The research will provide a solution to these drawbacks by stating an intelligent cyber threat detection framework that would be built using an optimized deep neural network in an IoT social network setting. First, an extensive data preprocessing and normalization process is implemented to improve quality and consistency of records of IoT traffic. After that, Improved Gorilla Troops Optimization (IGTO) is used with the help of a Decision Tree approach to select the best features to exclude the redundant and irrelevant attributes, but not to lose discriminative information. The chosen features are then categorized with the help of a deep neural network the parameters of which are optimally adjusted with the Harris Hawks Optimization (HHO) algorithm to enhance the convergence and classification. The suggested framework is tested with the CICIoT2023 data that includes various benign and malicious IoT traffic case scenarios. As experimental data shows, the given model leads to greatly enhanced detection accuracy, precision, recall, and F1 Score with lowered false-positive rates as compared to the current deep learning-based methods. The results prove that the combination of bio-inspired optimization methods and deep neural networks presents a powerful and scalable approach to intelligent cyber threat detection in dynamic environments of IoT social networks