Information-Theoretic Medical Image Encryption via LLE-Verified Chaotic Keystreams and DNA Diffusion

Area of Science:

• Cryptography and Information Security
• Applied Mathematics and Computational Science
• Artificial Intelligence and Machine Learning

Background:

• Traditional encryption methods face evolving security threats.
• Need for robust keystream generation in secure communication.
• Leveraging chaotic systems and neural networks for enhanced security.

Purpose of the Study:

• To propose and evaluate a novel information-theoretic encryption scheme.
• To utilize a four-dimensional chaotic map and LSTM neural network for keystream generation.
• To analyze the security and performance of the proposed encryption scheme.

Main Methods:

• A four-dimensional chaotic map driver coupled with an LSTM neural network for keystream generation.
• Security analysis involving permutation phase for spatial redundancy removal and invertible DNA cross-diffusion.
• Evaluation using Shannon's entropy, pixel intercorrelation, and differential analysis (NPCR, UACI).
• Key space analysis and assessment of decryption validity.

Main Results:

• Keystream generation limited by largest Lyapunov exponent (LLE) test.
• High Shannon entropy (7.99-8.00 bits per channel), indicating low uncertainty.
• Effective decorrelation shown by low pixel intercorrelation and strong differential attack resistance (NPCR ≈ 99.6%, UACI ≈ 33.3%).
• Extensive key space (order ≈ 2^384) ensuring robust security and perfect decryptability.

Conclusions:

• The proposed encryption scheme offers strong security and high performance.
• Chaos verification significantly enhances the statistical properties of keystreams.
• The combination of chaotic maps and LSTM networks provides a promising approach for modern encryption.