Hybrid Architecture for the Detection of AI-Generated Synthetic Images
8th International Congress on Human-Computer Interaction, Optimization and Robotic Applications, ICHORA 2026, Ankara, Türkiye, 21 - 23 Mayıs 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/ichora69329.2026.11537048
- Basıldığı Şehir: Ankara
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: AI-generated images, Convolutional Neural Networks (CNN), generative AI, image classification, latent diffusion, Vision Transformer (ViT)
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Artificial Intelligence (AI) generated images have attracted significant attention in recent years due to the rapid development of generative AI and Latent Diffusion Models (LDMs). Therefore, identifying such generated images is crucial for both ethical and legal reasons. Nevertheless, accurately performing this identification remains a difficult problem. This paper presents a hybrid deep learning model designed to distinguish authentic images from AI-generated counterparts using the CIFAKE dataset. The architecture of the model is combination of DenseNetbased Convolutional Neural Network (CNN) backbone and Vision Transformer (ViT) blocks, thereby utilizing the complementary properties of both paradigms for improved image classification performance. Experimental findings indicate that the proposed hybrid model attains an accuracy of 96.77 % and an F1-score of 96.77 %, surpassing the baseline model as well as the other models reported in the literature. The results clearly point to the fact that deep learning models should focus not only on microscopic 'digital fingerprints' such as pixel-level noise but also on larger structural relationships within an image to achieve reliable detection.