Transfer learning with XAI for robust malware and IoT network security.

Area of Science:

• Cybersecurity
• Machine Learning
• Digital Forensics

Background:

• Increasing privacy-exploiting malware necessitates advanced detection methods.
• Obfuscation techniques make real-time malware detection challenging.
• Traditional forensic analysis requires sophisticated pattern identification.

Purpose of the Study:

• Develop a deep learning model for classifying obfuscated malware.
• Enhance intrusion detection in IoT and network traffic using transfer learning.
• Improve transparency and interoperability of malware detection models.

Main Methods:

• Developed a deep learning model trained on the MalwareMemoryDump dataset.
• Implemented transfer learning to adapt the model for NF-TON-IoT and UNSW-NB15 datasets.
• Integrated Explainable AI (XAI) for model transparency.

Main Results:

• Achieved 99.9% accuracy on MalwareMemoryDump and 96% on NF-TON-IoT and UNSW-NB15 datasets.
• Demonstrated improved accuracy and efficiency in cross-domain detection.
• Showcased reduced training time and computational costs via transfer learning.

Conclusions:

• The proposed deep learning model effectively handles diverse cybersecurity threats.
• Transfer learning significantly enhances malware detection across different domains.
• The model offers a highly effective and generalizable approach outperforming existing security techniques.