Deep Learning-Driven Behavioral Analysis for Real-Time Threat Detection and Classification in Network Traffic

With the evolution of digital spaces, cyber threats now evolve to more complex forms, requiring innovative solutions for real-time intrusion detection and classification for network traffic. Cybersecurity is also critical for building resilient infrastructure, which is one of the goals of the United Nations, which emphasizes secure and sustainable digital ecosystems. This research proposes a framework powered by deep learning that employs an enhanced fully connected neural network (EFNN) to analyze behavior and detect threats. The proposed algorithm, Enhanced Fully Connected Neural Network-Based Threat Detection (EFNN-TD), fuses advanced data preprocessing with FCBF-based feature selection and SMOTE-based handling of class imbalance. We introduce a novel multi-featured attentive framework for generating practical and compact representations for multi-class classification in anomaly correlation. It identifies and classifies various network intrusions efficiently with precision and recall, which are required to reduce false alarms and detect all threats. The proposed system can aid in safeguarding digital infrastructures through real-time monitoring and decision-making, thus supporting the global imperative for promoting secure, robust, and sustainable technological advancement to foster economic growth and societal development.