Deepfake Image Detection: A Review of Existing Methods and a Hybrid CNN-Based Proposed Framework

The rapid advancement of generative adversarial networks (GANs) and related synthesis techniques has enabled the creation of highly realistic deepfake images, posing significant risks to security, privacy, and trust in digital media. Existing detection methods range from handcrafted forensic features to deep learning-based models, yet they often face challenges such as dataset bias, limited cross-domain generalization, and vulnerability to adversarial manipulations. This paper reviews state-of-the-art methods for deepfake and image forgery detection, highlighting critical research gaps, particularly in robust feature representation and scalability. To address these limitations, we propose a hybrid convolutional neural network (CNN)-based framework that integrates block-based ResNet-50 for effective feature extraction and VGG-16 for classification. Publicly available datasets such as Celeb-DF and the DeepFake Detection Challenge (DFDC) are discussed alongside challenges of real-world deployment. The study concludes by outlining future research directions, including multimodal detection, continual learning, and explainable AI for improved interpretability.