ClearShot uses motion tracking software to read what your hands are typing
Last Updated on Wednesday, 21 May 2008 08:34 Written by admin Wednesday, 21 May 2008 08:34
Inspired by the Robert Redford flick “Sneakers” (1992), a group of researchers at the University of California – Santa Barbara have developed a process they call “ClearShot” that uses a variety of novel techniques including motion tracking, sentence reconstruction, and error correction to reconstruct a typed message or password.
The popularity of Web cams today partly inspired this research since a Web cam exploit is a relatively simple hack. The research is fascinating to read in its depth of coverage of related technology and its examination of the complexity of this challenge. Other forms of spying might include video surveillance from a cell phone or a more distant (and sophisticated) video device.
From the published research:
“Note that, differently from the situation represented in “Sneakers,” we are not interested in the recovery of a single password (which could probably be easily recovered by a human being). Instead, our goal is to recover all the information entered by the user through the keyboard (e.g., email messages, instant messages, documents, and code). This type of activity would be error-prone and resource-intensive if performed by a human being analyzing manually each of the frames produced by the camera. Our experiments show that, for a human, reconstructing a few sentences requires lengthy hours of slow-motion analysis of the video.
“Automatically recognizing the keys being pressed by a user is a hard problem that requires sophisticated motion analysis. Previouswork in the computer vision field has produced algorithms that can perform well only when the user types on the keyboard using non-realistic movements (e.g., one finger at a time [9, 47]). In addition, several “enhanced” typing interfaces (e.g., projected keyboards) use an array of non-visual sensors to identify the keys being pressed. Our approach is different because it aims at reconstructing the typed information in a realistic setting and using exclusively the video information captured by an off-the-shelfweb cam.”
I think that my barely adequate hunt-hunt-hunt some more and peck typing will be able to defeat it.
thanks nice post