Diabetic Retinopathy (DR) is a leading cause of vision loss worldwide, primarily affecting individuals with prolonged diabetes. Early detection is crucial to prevent irreversible blindness. This project proposes a secure and intelligent framework for the early detection of Diabetic Retinopathy using Federated Learning (FL), ensuring both data privacy and efficient model training across multiple healthcare institutions. The system utilizes Optical Coherence Tomography (OCT) images for accurate retinal analysis and employs Convolutional Neural Networks (CNNs) such as ResNet-50 and VGG-16 for disease classification. To enhance data security, Multi-Factor Authentication (MFA) is integrated, allowing only authorized medical professionals to access sensitive information. Unlike traditional centralized AI models, the proposed system prevents raw data sharing, thus maintaining patient confidentiality while improving diagnostic performance. Future enhancements, including Homomorphic Encryption (HE) and Explainable AI (XAI), will further strengthen data protection and interpretability of results. Overall, this system contributes to Sustainable Development Goal (SDG 3): Good Health and Well-being, by promoting accessible, secure, and intelligent healthcare solutions for early eye disease diagnosis.