Year: 2026
Type: article
Abstract: This research proposes a dual-domain image encryption framework for securing color images.The system creates a dynamic encryption key through the Convolutional Neural Network (CNN) and optimizes it via the Particle Swarm Optimization (PSO) algorithm.The primary key is split into two parts, one for spatial-domain encryption (K_spat) and one for frequency-domain encryption (K_freq).In the spatial domain, permutation and XOR operations are performed, and in the frequency domain, mid-and high-frequency components are encrypted using the subkey K_freq after Discrete Wavelet Transform (DWT).The encrypted image is reconstructed through Inverse DWT (IDWT).The system was tested on security and quality measures including entropy, correlation coefficient, NPCR, UACI, and histogram analysis.Experimental results revealed entropy values from 7.992 to 7.998, correlation coefficient less than 0.00002, NPCR 99.58%, and UACI 33.45%, demonstrating high resistance to statistical, differential, noise, and cropping attacks.Dynamic key generation guarantees unique keys for each transmission.CNN models showed more than 99.75% classification accuracy.In conclusion, the CNN-based feature extraction, PSO key optimization, and dual-domain DWT encryption propose a smart, dynamic, and secure image encryption.
Institution: University of Basrah
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7140040367