A software application could effectively warn users when they are about to give away sensitive personal information online.
The software, originally developed to adapt eye recognition equipment for use in their behavioral research involving online information disclosure, is showing promise in displaying warnings in a dynamic manner that is more readily perceived and less easily dismissed by the user, said Dr. Frank Zhu, a computer science associate professor at the University of Alabaman in Huntsville, and Dr. Sandra Carpenter, a psychology professor.
Computer Science doctoral student Mini Zeng has been working on the software and the behavioral research for about 2 ½ years.
“I need to know how long the user’s eyes stay on the area and I need to use that input in my research,” Ming said.
The eye tracker detects where a user’s eyes are at the computer screen and records how long they gazed at that spot. Zeng uses these two functions to find when a user’s eyes remain on a request for sensitive personal information. At that moment, a warning box displays. The app tracks the amount of time the user’s eyes are on the warning, and the box stays on the screen until sufficient time has passed to ensure it has been read, then when the user looks away it disappears.
“That’s the novelty here, it is using the eye tracker as an input to warn people what not to do,” Zhu said.
If the user looks away from the warning, it remains active until the app detects enough time has been spent on it to read it.
The relative unpredictability of a warning that can pop up anywhere on a screen when a user is looking at a request to divulge personal information helps overcome behavioral obstacles to paying attention to standard warnings the researchers have identified in their work.
“If you get a warning every single time and it becomes annoying or habitual, you are going to ignore it,” Carpenter said.
For their behavioral research, Zeng created an app that mimics a restaurant reservation app asking for personal information. That app ends up used along with the warning software to determine the effectiveness of warnings in test subjects.