Enhancing malware detection and classification in network traffic using deep learning techniques.

Area of Science:

• Cybersecurity
• Network Security
• Machine Learning

Background:

• Traditional malware detection struggles with complex, evolving network traffic.
• High false positives and difficulty in real-time detection are key challenges.
• Need for advanced methods to identify malicious activities accurately.

Purpose of the Study:

• To enhance malware detection accuracy and reduce false positives in network traffic.
• To enable real-time malware detection in dynamic network environments using deep learning.
• To investigate advanced deep learning techniques for improved cybersecurity.

Main Methods:

• Entropy-Based Traffic Filtering (ETF) for anomaly identification.
• Self-Supervised Learning for Anomaly Detection (SSLAD) for unlabeled data.
• Graph Neural Networks (GNN-MTC) and Context-Aware Graph Attention Networks (CA-GAT) for relational traffic analysis.

Main Results:

• The proposed deep learning model achieved 98% accuracy.
• Outperformed existing methods like DeepMAL (95%) and Huang et al. (97.3%).
• Demonstrated superior precision and recall for detecting known and novel malware.

Conclusions:

• The deep learning model is highly effective for real-time network security applications.
• The approach shows strong performance in detecting diverse and emerging malware threats.
• Future work includes adaptive learning and hybrid architectures for enhanced robustness.