• Manimozhi. G
• Sukhvinder Singh

Nowadays Iris attacks have become a common thing, even though the iris has unique features to differentiate each person, hackers manipulate the biometric authentication system using various types of fake iris images such as printed iris images, contact lens images, patterned iris images, and cadaver iris images. There are a number of techniques or models that have already been proposed for Iris Liveness Detection using Deep Learning Networks. In this paper, a three-class scenario is suggested that uses a novel serial architecture for Presentation Attack Detection based on MobileNetV3 and EfficientNetB0 to improve the efficiency, and accuracy and reduce the relevant error rates of the system by making it lighter. The MobileNetV3 and EfficientNet will be trained from scratch to distinguish between presentation attack images and genuine iris images. The bonafide class comprises live iris pictures, while the assault show instrument class comprises of printed, and patterned riris pictures for three species. The result from the MobileNetV3 and EfficientNetB0 will be compared with other existing deep learning networks such as VGG-19 and MobileNetV2. From the experimental results and analysis, we can see that the MobileNetV3 performs better than the other deep learning networks by providing improved accuracy, efficiency with less time. The proposed models gives out 0.24% APCER and 0.12% BPCER for MobileNetV3 and 0.38% APCER and 0.23% BPCER for EfficientNet. MobileNetV3 takes 19 milliseconds for each step whereas EfficientNet takes 34 milliseconds and VGG-19 takes 85 milliseconds.

Biometric authentication system, Iris Liveness Detection, Presentation Attack Detection, deep learning networks, confusion matrix.

"Iris Liveness Detection using MobileNetV3 and EfficientNet using LivDet Dataset", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 3, page no.339-349, March-2019, Available :http://www.jetir.org/papers/JETIR1903O50.pdf

"Iris Liveness Detection using MobileNetV3 and EfficientNet using LivDet Dataset", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 3, page no. pp339-349, March-2019, Available at : http://www.jetir.org/papers/JETIR1903O50.pdf