An Intelligent Machine Learning Framework for Cybersecurity Threat Detection and Monitoring

Abstract

As the number of IoT devices increases rapidly, cybersecurity threats are becoming more dynamic and intricate, necessitating more sophisticated and intelligent defense mechanisms. This work presents a unique intrusion detection system that enhances the security of the IoT by combining signature-based and anomaly-based techniques with AI-driven threat mitigation. The proposed framework is an amalgamation of Recurrent Neural Network (RNN) and Extreme Gradient Boosting (XGBoost) models that efficiently process network traffic, detect suspicious activities, and allow a real-time response to threats. According to the ToN-IoT dataset, data cleaning, normalization, and data balancing are employed to improve performance. The experimental analysis indicates the high-performance rate where the RNN and XGBoost have 98.5% and 98.7% accuracy respectively. The RNN model exhibits a marginally greater trade-off in terms of precision, recall and F1-score metrics. The efficacy of the proposed approach in the process of achieving high detection performance is validated by the comparative analysis with the currently existing models such as Naive Bayes, LSTM, CNN, and Random Forest. The research as a whole indicates the potential of DL and ensemble methods to improve the process of cybersecurity threat detection in the IoT context, as well as to address the issues of data imbalance and feature relevance.