Embedded Files
Iris Spoofing and Presentation Attack Detection
Iris Spoofing and Presentation Attack Detection
Students: Vera Casquero García, Zelalem Haile
Research mentor: Rasel Ahmed Bhuiyan
Biometric security systems are increasingly used for authentication. Iris recognition is one of the most secure biometric methods due to the uniqueness of the iris patterns. However, it remains vulnerable to attacks such as printouts, digital images, contact lenses, and post-mortem irises. Presentation Attack Detection (PAD) is crucial for enhancing biometric security by identifying such attacks, ensuring user convenience, and safeguarding against unauthorized access. To explore the vulnerabilities of iris recognition systems to replay attacks, we curated a dataset comprising 2651 bonafide (genuine) iris images sourced from the ND3D dataset. The printed replicas of these bonafide iris images were collected using the IriTech sensor. We split this generated dataset into three subsets: train, test, and validation, and trained a ResNet-18 model for iris PAD. The model achieved perfect detection performance on the test data, with accuracy, precision, recall, and F1 scores all at 100%, along with an area under the ROC curve of 100%. The model also recorded a BPCER (Bona Fide Presentation Classification Error Rate) of 0.00% and an APCER (Attack Presentation Classification Error Rate) of 0.00%. These results suggest that the simplicity of the dataset contributed to the model's high performance.
Rasel Ahmed Bhuiyan received a B.Sc. degree in Computer Science and Engineering (CSE) from the University of Asia Pacific (UAP), Dhaka, Bangladesh, in 2018. From 2018 to 2021, he served as a Lecturer in the CSE department at Uttara University, Dhaka, and was actively engaged in teaching.
Currently, he is pursuing his Ph.D. in Computer Engineering at the University of Notre Dame and working as a Research Assistant with the Computer Vision Research Laboratory (CVRL) under the supervision of Prof. Adam Czajka. His research interests revolve around computer vision techniques applied to iris biometric technologies and developing generalized machine learning models to address challenges in iris biometrics. He has contributed to several research publications at the intersection of machine learning, computer vision, and pattern recognition.
Dr Adam Czajka's research focuses on human biometrics and security, especially on iris recognition and methods of detecting unknown presentation attacks. In his research he tries to understand how human perception of abnormal signals can be effectively used to build computer algorithms generalizing better on unknown spoofs and how computer algorithms can aid humans in examining biometric samples.
In general, he is fascinated by a wide spectrum of research in computer vision, pattern recognition and machine learning, and the non-obvious intersections with psychology, medical sciences, and art.
Page updated
Report abuse