Iris recognition outlooks foremost biometric identification technique, celebrated for its accuracy and steadiness. This paper proposes an improved deep learning framework through optimal Gabor filters and deep belief network (DBN), fused with gated recurrent units (GRU). The method begins with Gabor filters that elicit key features from iris visuals, followed by DBN for structured feature learning. GRUs are then used to extract the sequential insights in iris images, improving the system's ability to manage disparities in image acquirement conditions and occlusions. Investigational assessments on benchmark iris images validate that the proposed architecture distinctly increases recognition precision and resilience to variations in lighting, occlusion, and noise. The outcomes indicate that the blend of Gabor filters, DBN, and GRU efficiently addresses the limitations of traditional iris recognition methods. This improved technique shows significant ability for biometric validation and security uses, maintaining high accuracy in stimulating environments, making it compatible for practical security developments.