التحديات والحلول في الكشف عن الصور المزيفة: نموذج يعتمد على الذكاء الاصطناعي والشبكات العصبية الالتفافية (CNN)

Abstract

Thanks to the progress of IT technologies, fake images are increasingly popular especially those produced with the help of modern tools like Deepfake. The goal of this work is to build a model using the CNN to identify fake images with a high level of accuracy as well as speed. The study employed an open dataset of real and fake images, which captured different patterns and complexity.

The proposed model was intended to have five main layers of convolutional, pooling, and classification layers. It was trained with Adam Optimizer and Binary Cross-Entropy loss function. The outcomes showed that the proposed model had an accuracy of 94.7 percent, which was far better than conventional models, such as SVM, that needed manual feature engineering. The ROC curve also confirmed this by giving the model AUC of 0.96, which shows how well the model performs in differentiating the real and

fake images.

The study concludes that CNNs are a very effective method of identifying fake images. Suggested improvements are increasing the size of the training data set and using superior methods to enhance the performance with more elaborate fake images. This model can be implemented practically in media and security in order to increase the

reliability of digital images given the new challenges posed by new digital forgery techniques.