The rapid growth of biometric authentication systems has raised concerns about their vulnerability to spoofing attacks, especially in Iris recognition. This thesis proposes a novel approach to Iris Liveness Detection using Convolutional Neural Networks (CNNs) to identify differences between authentic and fraudulent Iris images, enhancing the security of these systems. A custom dataset was created, featuring authentic images of Iris from five individuals, taken at various angles, along with fake images synthesized from their photographs. The CNN model, trained on this dataset, achieved an impressive 92.51% accuracy. Its robustness was tested on unseen data under diverse conditions, demonstrating strong generalization potential. The thesis also explores the model's interpretability, aiding in refining the liveness detection mechanism to counter evolving spoofing techniques effectively.