Detecting malicious code variants using convolutional neural network (CNN) with transfer learning.

Area of Science:

• Computer Science
• Cybersecurity
• Network Security

Background:

• Malware poses a significant threat to digital systems, necessitating advanced detection methods.
• Current anti-malware solutions and detection techniques often struggle with efficiency and accuracy.
• Existing image-based malware detection using grayscale images is computationally intensive.

Purpose of the Study:

• To develop a novel and efficient malware detection system.
• To improve the accuracy of malware variant classification.
• To overcome the limitations of grayscale image conversion for malware analysis.

Main Methods:

• Proposed the Malware Variants Detection System (MVDS).
• Transformed malicious code into color images for analysis.
• Utilized transfer learning for automated malware image classification.

Main Results:

• Achieved a classification accuracy of 97.98%.
• Demonstrated high detection speed compared to traditional methods.
• Color image conversion proved more effective than grayscale.

Conclusions:

• The MVDS offers enhanced malware detection capabilities.
• The system's high accuracy and speed make it suitable for practical network security.
• Leveraging color images and transfer learning significantly improves malware identification.