Real Time Threat Detection for Critical Infrastructure using Machine Learning Techniques

Abstract

The growing reliance on telecommunications as a critical infrastructure has intensified the demand for advanced cyber threat detection systems. Traditional security tools often fail to address the speed, volume, and complexity of modern attacks, especially in dynamic telecom environments. This study presents the design and deployment of a real-time threat detection framework powered by Artificial Intelligence (AI). The system combines a fine-tuned BERT-based deep learning classifier with a heuristic rule-based fallback engine to achieve both high accuracy and interpretability. Using the CICIDS2017 dataset for training and the UNSW-NB15 dataset for evaluation, the framework processes network traffic through structured preprocessing and binary classification. It is deployed using a Flask-based backend and a responsive frontend that displays real-time alerts and performance metrics such as accuracy, precision, recall, and F1-score. Results show that the system reliably detects multiple categories of cyber threats with low latency and supports operational scalability. The research demonstrates a practical and intelligent approach to securing telecom networks, with future work aimed at integrating live data streams and enhancing model transparency through explainable AI.