LSTM-driven chaotic keystream generator for robust medical image encryption.

Area of Science:

• Medical Imaging
• Cybersecurity
• Applied Mathematics

Background:

• Digital healthcare and telemedicine necessitate secure medical image transmission and storage.
• Conventional encryption methods struggle with large medical image data volumes and pixel correlations.
• Existing chaos-based encryption has limitations in key space and periodicity due to finite precision.

Purpose of the Study:

• To develop a novel, secure, and efficient hybrid encryption scheme for medical images.
• To address the limitations of conventional and existing chaos-based encryption methods.
• To improve the robustness and key space of image encryption techniques.

Main Methods:

• A hybrid encryption scheme integrating Long Short-Term Memory (LSTM) networks with chaotic maps.
• Generation of adaptive keystreams using a hash float-seeded chaotic map and statistical features.
• Experimentation with three variants: LSTM+Logistic, LSTM+sine map, and LSTM+Chebyshev map.

Main Results:

• All three hybrid models demonstrated strong encryption performance.
• The LSTM+Logistic map variant showed superior performance in randomness, noise, and crop attack resistance.
• Achieved high NPCR (99.62%), UACI (39.59%), and entropy (7.975).

Conclusions:

• The proposed hybrid encryption method offers enhanced security and computational efficiency for medical images.
• The LSTM+Logistic map variant provides a robust solution for secure medical image transmission.
• The method is suitable for practical deployment in digital healthcare systems.