Ensuring Secured Iris Authentication for Mobile Devices

摘要

The secured authentication in mobile devices and individual consumer products based on iris patterns is a vital challenge. An adversary may deceive the IR system through artifacts such as printed iris, synthetic eyeballs, textured contact lenses, etc. This paper introduces an approach to discriminate between the real and forged iris, through the pooling of multi-scale BSIF and Fine-tuned DenseNet features extracted from raw and normalized iris. The paper also validates the generalizability of the proposed approach across several attack variants, iris sensors, and datasets. Besides, the experiments performed on the publicly available LivDet-2017 datasets endorse the efficacy of the proposed method.