Communications of the ACM
Forget Passwords, a Student Research Team Wants You to Log in By Speaking a Sentence
Credit: DeviantArt
Student researchers at California State University, Fullerton (CSUF) are developing new biometric security technologies, including those based on facial recognition, fingerprints, and heartbeat sensors. The team, led by CSUF professor Mikhail Gofman, wants to enhance mobile phone and tablet security by creating a multimodal framework, which means using multiple inputs, such as face and voice, to unlock or secure mobile devices.
The researchers say improved biometric authentications could be a more secure authentication protocol than traditional methods such as passwords or pins, and they note conventional sensor technology and biometric algorithms are not robust enough to resist sophisticated spoofing attacks or do not work well in less than ideal conditions. "We believe that a multimodal approach, which viably consolidates features from multiple biometric modalities, e.g., face and voice features, can address many of these problems," Gofman says.
The CSUF approach has 20-percent error rates, compared to 40-percent error rates found in traditional algorithms commonly used in face and voice recognition. "Our preliminary results indicate that such 'multimodal' fusion yields significantly more accurate authentication results than methods that use only one single biometric trait," says CSUF professor Sinjini Mitra.
The CSUF prototype enables users to record a short video of their face while speaking a sentence, and the samples are statistically fused to create a third level of integrated biometrical inputs.
From Daily Titan
View Full Article
Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA
No entries found