PAFE: A lightweight visualization-based fast malware classification method.

Area of Science:

• Cybersecurity
• Computer Science
• Machine Learning

Background:

• Automated malware toolkits present evolving cybersecurity threats.
• Existing visualization-based malware analysis methods struggle with texture feature variations and require deep networks, sacrificing speed.
• Challenges include preprocessing alterations and class imbalance in small datasets.

Purpose of the Study:

• To propose PAFE, a lightweight and rapid visualization-based malware classification method.
• To address texture feature variations and class imbalance issues in malware analysis.
• To improve both efficiency and effectiveness in malware variant classification.

Main Methods:

• Pixel-filling techniques to handle texture feature variations during preprocessing.
• Data augmentation to manage class imbalance in small sample datasets.
• Multi-scale feature fusion and channel attention mechanism within a modular design for enhanced feature expression.

Main Results:

• PAFE achieves a 99.25% accuracy rate for malware variant classification.
• The method demonstrates a rapid prediction time of 10.04 ms.
• PAFE outperforms current state-of-the-art methods in both efficiency and effectiveness.

Conclusions:

• PAFE offers a significant advancement in rapid and accurate malware classification.
• The proposed method effectively overcomes limitations of existing visualization-based approaches.
• PAFE provides a promising solution for real-time cybersecurity threat detection.