Optimization of a Deep Learning Algorithm for Security Protection of Big Data from Video Images

Area of Science:

• Computer Science
• Cryptography
• Deep Learning

Background:

• Digital technology proliferation increases data security risks.
• Cloud platforms require robust protection for image and video data.
• Existing encryption methods face challenges from advanced cyber threats.

Purpose of the Study:

• To develop a deep learning-based encryption algorithm for secure cloud storage and sharing of image and video data.
• To enhance privacy protection against neural cryptography and sophisticated attacks.
• To improve the security of existing neural network encryption algorithms.

Main Methods:

• Image saliency detection to identify important regions in video images.
• Adaptive division and reorganization of important and non-important image regions for encryption.
• Development of an improved encryption algorithm based on selective ciphertext attack analysis.
• Comparative analysis of algorithm security capabilities.

Main Results:

• The proposed algorithm reduces decryption error rates over time.
• Attacker's (Eve) classification error rate increases, with accuracy not exceeding random prediction when secure networks are learned.
• Chosen ciphertext attack-advantageous neural cryptography (CCA-ANC) demonstrates efficient performance (14s encryption, 69mb/s speed).

Conclusions:

• The self-learning secure encryption algorithm significantly enhances password security and data protection for video images.
• The method provides a robust defense against advanced cyber threats in cloud environments.
• The algorithm offers a practical and efficient solution for securing sensitive visual data.