*Result*: PriFL-XAI: Hybrid privacy-preserving federated learning models for monkeypox detection through GAN augmentation and explainable AI.

*Monkeypox has emerged as a critical global health concern, demanding timely and precise diagnostic measures to control its spread. Despite its similarity to other pox viruses, accurate detection is hindered by the scarcity of privacy-protected medical imaging data, which complicates the application of conventional deep learning methods. In response, this paper introduces a secure and privacy-preserving framework that combines generative adversarial networks (GANs) for data augmentation with explainable artificial intelligence (XAI) for enhanced interpretability. Using StyleGAN2_ADA, we synthesize a wide range of image samples to address data imbalance, thus improving the robustness of the training process. We further merge established convolutional neural networks such as ResNet18, ResNet50, MobileNetV3, and DenseNet121 with Vision Transformers (ViT) to create a suite of hybrid architectures. These models operate within a Flower-based federated learning ecosystem, enabling large-scale collaborative training without exposing sensitive patient information. In addition, XAI methods provide actionable insights into the decision-making of each model, thereby fostering transparency and clinical trust. Empirical tests show that the hybrid DenseNet121–ViT-B32 model achieves 98.2% precision, 95.25% recall, and 96% F1 Score, thus validating the prospects of our federated, GAN-enhanced, and XAI-supported solution for precise monkeypox detection. This framework demonstrates real-world applicability by enabling secure and scalable diagnosis, with GAN-based augmentation increasing dataset size by over 700%, significantly boosting model generalization. The research underscores the need to balance privacy-centric interventions and explainable deep models in the treatment of emerging infectious diseases and the protection of global public health. [ABSTRACT FROM AUTHOR]*