<![CDATA[The rapid progress in artificial intelligence has enabled the creation of highly realistic images, leading to concerns about the credibility and genuineness of visual content. This study aims to address the growing challenge of verifying the authenticity of digital images in the era of advanced generative artificial intelligence by developing an effective method to detect AI-generated (deepfake) images. To achieve this objective, we employed a Machine Learning (ML) framework based on Convolutional Neural Networks (CNNs), evaluating five established architectures, VGG19, ResNet50, Xception, DenseNet-121, and InceptionV3, through a systematic pipeline involving dataset compilation, image preprocessing, feature extraction, model training, and rigorous validation. Our experimental analysis demonstrates that DenseNet-121 and InceptionV3 achieve state-of-the-art performance, both attaining 98% accuracy in distinguishing AI-synthesized images from real ones, despite a non-negligible error rate observed in other models. These findings highlight the viability of CNN-based approaches for reliable deepfake detection. The novelty of this work lies in its comparative assessment of multiple CNN architectures on a curated dataset of AI-generated imagery, offering practical insights into model selection for forensic and security applications. The proposed method contributes a robust, scalable solution with significant implications for digital content moderation, cybersecurity, and multimedia forensics, where timely and accurate identification of synthetic media is increasingly critical.]]>
Copyrights © 2026
