Generation of Fake Iris Images Using CycleGAN

With the development of biometric recognition technology, identification of users through biometric information such as iris, fingerprint, and palm print is being applied to many areas. In the case of iris, various methods of recognition and methods of detection of fake iris have been studied at the iris recognition stage. However, fake iris detection research has been conducted by using the printed output or the artificial iris due to the absence of fake iris data. In this paper, fake iris images are generated for the research of the detection of fake iris using CycleGAN. The CycleGAN model has learned to reduce constraints against existing generation models and to avoid bias in probability distributions using bidirectional LossFunction. In the experiment, CASIA Iris Image Database ver 4.0 was used and the data was obtained for the detection of fake iris by creating fake iris.