E2E-RDS: Efficient End-to-End Ransomware Detection System Based on Static-Based ML and Vision-Based DL Approaches.

Area of Science:

• Cybersecurity
• Machine Learning
• Computer Vision

Background:

• Ransomware poses a critical cyber-malware threat, necessitating advanced detection methods.
• Existing malware detection approaches face challenges due to innovative evasion techniques by attackers.

Purpose of the Study:

• To introduce an efficient End-to-End Ransomware Detection System (E2E-RDS).
• To comprehensively utilize and compare static-based and vision-based ransomware detection (RD) approaches.

Main Methods:

• Static-based RD: Reverse engineering code to extract features for ML models.
• Vision-based RD: Converting executables to images for analysis by Convolutional Neural Network (CNN) models.
• Leveraging Fine-Tuning (FT) and Transfer Learning (TL) in CNN models for enhanced detection.

Main Results:

• Static-based RD with Ada Boost (AB) achieved 97% accuracy.
• Vision-based RD with FT ResNet50 CNN reached 99.5% accuracy.
• Vision-based approach demonstrated superior efficiency and cost-effectiveness by avoiding feature engineering.

Conclusions:

• The E2E-RDS is a versatile and highly efficient solution for ransomware detection.
• The vision-based RD approach is more powerful and cost-effective than static-based methods.
• E2E-RDS shows promise for real-time ransomware detection across various systems.