This paper introduced a novel Iris Liveness Detection system based on EfficientNet-B7 combined with ISO/IEC 29794-6:2015 quality metrics such as Iris Pupil Concentricity (IPC), Iris Radius, and Iris-Pupil Contrast to improve biometric authentication. Conventional iris recognition systems are susceptible to attacks like printed images, videos or fake eyes, resulting in security breach. Existing methods often fail to effectively generalize across diverse attack types and suffer from high error rates. To address these limitations, we employ a deep learning-based approach that incorporates image quality assessment for enhanced robustness. The model is trained and evaluated on the Clarkson and CASIA datasets, utilizing metrics such as Accuracy, Precision, Recall, F1-Score, Specificity, APCER, and BPCER. The proposed method achieves accuracy 99% & 96.99, precision 99.6% & 96.7, recall 99.2% and 97.3%, and specificity 98.9% & 96.7%, with an F1-score of 99.5% for Clarkson and CASIA datasets respectively. Additionally, the system demonstrates a low APCER (1 & 3.3), BPCER (0.8 & 2.7), and ACER (0.9 & 3) for Clarkson and CASIA datasets respectively., indicating strong resilience against spoof attacks. The results confirm the efficacy of the proposed approach in distinguishing live irises from spoofed ones, thereby strengthening biometric security. This study demonstrates that integrating EfficientNet-B7 with iris quality metrics significantly improves liveness detection performance.