Image Encryption Algorithm Based on Convolutional Neural Networks and Dynamic S-Box Generation

2026-06-19 · Source: cs.SE updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, short

Summary

Submitted on June 18, 2026, a new image encryption algorithm proposes a dynamic approach by integrating Convolutional Neural Networks (CNNs) with classical cryptography. This method aims to enhance the security and flexibility of image encryption. The core innovation involves generating adaptive Substitution boxes (S-boxes) based on characteristics learned by a trained CNN. These CNN-based S-boxes offer increased non-linearity, uniqueness, and input image dependence compared to conventional fixed S-boxes, which are vulnerable to linear and differential attacks. This dynamic behavior strengthens the confusion property, making the algorithm more resistant to statistical and structural attacks. The encryption process includes CNN-based feature extraction and the creation of a personalized S-box for pixel replacement. Security assessments, utilizing metrics like Entropy, histogram analysis, correlation, NPCR, and UACI, indicate that this scheme is more resilient and flexible than traditional encryption methods.

Key takeaway

For AI Security Engineers evaluating robust image encryption, this research suggests that integrating Convolutional Neural Networks for dynamic S-box generation significantly enhances security. You should consider adopting adaptive S-box mechanisms to improve non-linearity and input dependence, making your systems more resilient against sophisticated statistical and structural attacks than conventional fixed S-box approaches. This method offers a flexible path to stronger visual data protection.

Key insights

A CNN-driven dynamic S-box generation method enhances image encryption security and flexibility against statistical and structural attacks.

Principles

• CNNs can generate adaptive cryptographic components.
• Dynamic S-boxes enhance non-linearity and input dependence.
• Hybrid CNN-classical crypto improves attack resilience.

Method

The method extracts image features using a CNN, then generates a personalized S-box based on these features to replace pixels. Security is assessed using Entropy, histogram, correlation, NPCR, and UACI metrics.

In practice

• Secure sensitive image data transmission.
• Protect visual data from statistical attacks.
• Develop adaptive cryptographic systems.

Topics

• Image Encryption
• Convolutional Neural Networks
• Dynamic S-box
• Cryptographic Security
• Statistical Attacks
• S-box Generation

Best for: Computer Vision Engineer, Research Scientist, AI Scientist, AI Security Engineer, Machine Learning Engineer

Related on AIssential

• An Improved CNN-LSTM Based Intrusion Detection System for IoT Networks — An improved CNN-LSTM model enhances IoT intrusion detection by integrating multi-class classification and temporal feature learning, achieving 97% accuracy.
• SNNF: An SNN-based Near-Sensor Noise Filter for Dynamic Vision Sensors — SNNF efficiently filters DVS noise using a compact binary representation and a hardware-optimized Spiking Neural Network.
• Masked Autoencoder for Image Reconstruction using Vision Transformers (PyTorch) — Masked Autoencoders with Vision Transformers learn robust image representations by reconstructing heavily masked image patches.
• Quality-Preserving Imperceptible Adversarial Attack on Skeleton-based Human Action Recognition — Adversarial attacks on S-HAR can be imperceptible by preserving motion quality through distribution-based optimization.
• An Improved CNN-LSTM Based Intrusion Detection System for IoT Networks — Hybrid CNN-LSTM models effectively capture both spatial and temporal features for robust IoT intrusion detection.
• Transformers Can Overcome the Curse of Dimensionality: A Theoretical Study from an Approximation Perspective — Transformers can overcome the curse of dimensionality, demonstrating strong expressive capabilities for function approximation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.SE updates on arXiv.org.