(Here’s a sample from my UAT days.)
Students Creating Multimodal Biometric System
Story by Trevor Green
Some computer laptops have fingerprint scanners that identify its users via distinctive ridge structures on their digits. The biometric security measure is one of many that hardware and software manufacturers employ to measure and analyze data to limit unauthorized access. A system that uses multiple forms of organic and digital protection, such as passwords and user behavior, would likely be a stronger security method.
University of Advancing Technology students Chase Schultz and Drew Porter are devising a multimodal biometric protection system that uses fingerprint scanning technology, password security and keyboard dynamics. The defensive triple-threat will recognize individuals by physical science, human memory and behavior-based patterns.
The goal is to prevent potential misuse of someone’s mobile computer. They hope to have a simple system that uses the security measures in tandem, devised in Linux and completely open source.
“What’s interesting is that the fingerprint scanner is just who you are; it identifies the person. And keyboard dynamics is more behavioral based, so it actually portrays something about the person. And then a password is something you know,” says Schultz.
He adds, “If someone sat down at your computer and tried chatting for you or something like that, it would boot them out.”
The pair was asked by Professor Shelley Keating to do a presentation on implementing a biometric system using more than one form of security protection. They chose Porter’s class emphasis on keyboard dynamics, a measurement of typing patterns (length of time and force to press keys), and Schultz’s fingerprint analysis work that uses open source software and fingerprint scanning hardware in Lenovo laptops. The use of passwords rounds out the program.
“So with that, all those measurements that we can use, we can use to make a biometric, which we can then use to make a security device and it just requires your keyboard and some software,” says Porter.
Schultz found open source libraries compatible with Lenovo’s fingerprint scanner co-processor. Porter did the same with keyboard dynamics software. Early tests of the program’s identification accuracy have shown it to be tough to crack. Porter notes a 95-percent rate to identify the user; the closest was Schultz at 88 percent.
“You can do the training for the system I was using up to 10 times, and after six times (which the more you do it the more accurate it gets) people are still not able to see who it was,” says Porter.
“I got pretty close to Drew’s but it’s just because I know how he types,” laughs Schultz.
The group plan to complete the program by the end of the fall semester. They are likely to register it on GetHub.com and use the website for feature requests and bug reports.