Machine learning-driven image encryption using SVM for enhanced security and computational efficiency

Area of Science:

• Computer Science
• Information Security
• Machine Learning

Background:

• Traditional image encryption methods face challenges in balancing security with computational efficiency, limiting real-time applications.
• Achieving high resilience often requires significant computational resources, hindering practical deployment.

Purpose of the Study:

• To develop a novel image encryption paradigm combining Machine Learning (ML) for improved space efficiency and security.
• To address the computational overhead associated with traditional image encryption techniques.

Main Methods:

• Utilized a Support Vector Machine (SVM) to classify image pixel blocks into low, moderate, and high information content.
• Implemented a selective encryption process encoding only moderate and high information blocks, leaving low information blocks unaltered.

Main Results:

• Achieved exceptional security evaluation metrics: 97.4% accuracy, 7.999 entropy, 0.0001 correlation, and 0.0153 energy.
• Demonstrated a considerable reduction in processing overhead through selective encryption.
• Validated the technique's effectiveness in preserving security while enhancing space efficiency.

Conclusions:

• The proposed ML-driven selective image encryption offers a unique, reproducible, and efficient solution for secure communications.
• This research advances image encryption by elaborating on computational aspects of ML-driven selective encryption.
• The method provides a viable approach for secure data transfer in complex modern communication systems.