Finding a suitable threshold value for an iris-based authentication system

Abstract

Authentication is the first line of defense of any information technology systems. One of the popular methods used today is biometric, and iris authentication is gaining popularity. However, the threshold value is deemed to be secure and appropriate has not been thoroughly studied. Threshold is a value that defines the acceptable amount of the correct bits of the image before securely passing the authentication process. Therefore, the main aim of this research was to find a secure and suitable threshold value used in iris authentication system, where iris localization was done by using Circle Hough Transform technique. Iris image databases v.4 from the Chinese Academy of Sciences Institute of Automatic (CASIA) were used in this research. The way to find the appropriate threshold was to test for the right balance of the GAR, FRR and FAR values when trying to verify the person’s identity. The results of the test revealed that the appropriate threshold had the value of 72.9246 percent of all the available bits of the iris image. Both had a high GAR and very low FAR and FRR values.  It can be concluded that the obtained threshold value was suitable and secure.