Ross receives best paper award at BTAS 2018The paper "DeepMasterPrints: Generating MasterPrints for Dictionary Attacks via Latent Variable Evolution" received the best paper award at the 9th IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS) that was held in Los Angeles last month. The paper was co-authored by Prof. Arun Ross along with colleagues from New York University - Philip Bontrager, Dr. Aditi Roy, Prof. Julian Togelius and Prof. Nasir Memon. In this paper, the author generate what are known as "MasterPrints" using a Generative Adversarial Network that employs a novel search algorithm referred to as Latent Variable Evolution. MasterPrints are fingerprints that can fortuitously match with a large number of fingerprints thereby undermining the security afforded by small fingerprint sensors. While previous work generated MasterPrints at the feature-level, this work generates MasterPrints at the image-level.
This work underscores the need to develop countermeasures for addressing dictionary attacks against biometric systems. It also validates the need for developing anti-spoofing technology for biometric systems, including those incorporated in smartphones and laptops. The work was supported by a grant from the National Science Foundation.
The paper can be accessed at https://www.cse.msu.edu/~rossarun/pubs/BontragerRossDeepMasterPrint_BTAS2018.pdf.
The conference website is at https://www.isi.edu/events/btas2018/home.
Arun Ross' previous work on this topic can be accessed at https://www.cse.msu.edu/~rossarun/pubs/RoyMemonRossMasterPrint_TIFS2017.pdf
(Date Posted: 2018-11-14)