Biometric Researchers Win Best Paper AwardThe paper "Spoofing PRNU Patterns of Iris Sensors while Preserving Iris Recognition”, co-authored by CSE PhD students Sudipta Banerjee and Vahid Mirjalili along with their advisor Prof. Arun Ross, received the Best Paper Award at the 5th IEEE International Conference on Identity, Security and Behavior Analysis (ISBA) held in Hyderabad, India in January 2019. This paper highlights the vulnerability of sensor identification algorithms which analyze an image and automatically determine the sensor that produced it. By judiciously modifying an iris image using an iterative scheme, the researchers demonstrate that it is possible to mislead sensor identification algorithms while still allowing the image to be used for biometric recognition purposes. This work has several practical implications. Firstly, it demonstrates the need to develop robust sensor identification algorithms that are not vulnerable to spoofing. Secondly, it can be used to extend privacy to biometric images such that information about the sensor that produced it (e.g., smartphone camera) can be masked. Thirdly, it motivates the need for developing robust digital forensic methods that can detect these type of adversarial attacks which undermine digital evidence in a court of law.
(Date Posted: 2019-02-05)