Enhanced Image Security by Integrating Discrete Cosine Transform and Convolutional Neural Network

As the need for image transmission with security becomes intensified, this study brings an enhanced encryption algorithm that combines Discrete Cosine Transform (DCT) and a Convolutional Neural Network (CNN) to improve efficiency and security. The proposed approach shows more complexity and unpredictability compared to the traditional chaotic encryption method with only logistic maps, in which DCT based noise and XOR based encryption are used. Further, contrasting image is enhanced with Contrast Limited Adaptive Histogram Equalization (CLAHE) prior to encryption to improve the quality of preprocessing for robust transformations. The proposed model employs DCT transformation in conjunction with CNN to strengthen encryption. This makes the algorithm stronger against cryptographic attacks. Decryption is achieved by applying XOR operations with DCT-generated noise, accurately recovering the original image. The combined use of deep learning-based encryption and DCT results in a more secure and complex model than traditional approaches. Furthermore, a user-friendly interface (UI) has been integrated to provide visualization of encryption results, improving accessibility and interpretability. A download feature is also introduced, enabling users to save encrypted images—an addition absent in previous implementations. Experimental evaluations confirm that this encryption framework offers improved security, computational efficiency, and user experience, establishing it as a robust solution for image encryption in secure communications.