Cyber Threat Detection Based on Artificial Neural Networks Using Event Profiles

Abstract

In the current digital era, cyber threats are be- coming increasingly complex, often targeting vital infrastructure, organizational networks, and personal computing environments. Traditional signature-based detection techniques frequently fall short in identifying emerging and zero-day threats. This research introduces a smart cyber threat detection framework that lever- ages Artificial Neural Networks (ANNs) and event profiling to address these limitations. The core challenge addressed is the inefficacy of conventional security mechanisms in recognizing unknown or novel intrusion behaviors. The proposed system employs ANNs to examine system and network event data—such as user activity logs, access attempts, and traffic patterns. The ANN is trained using a labeled dataset comprising both benign and malicious activity profiles. Feature extraction and normalization processes transform raw data into structured input vectors for the model, enabling the ANN to learn and differenti- ate between legitimate and anomalous behaviors. Experimental results demonstrate the ANN’s capability to perform real-time threat detection with over 94% accuracy, while significantly minimizing false positives. The system successfully detects a wide spectrum of cyber threats, including unauthorized access, suspicious traffic flows, and malware-like behavior. By integrating adaptive learning capabilities, this method not only improves detection precision but also evolves to recognize new threat vectors, offering a scalable and effective solution for modern cybersecurity infrastructures.