Have picture with out a name? A person you would like to find? Leave it to the new facial recognition tools being used by Google and Facebook to do your job. It is definitely convenient but to what extent could this blur the boundaries of individual privacy?
In a study held by Professor Alessandro Acquisti and his research team (including Ralph Gross and Fred Stutzman) the power facial recognition technology was put to the test. Three studies were held in Carnegie Mellon University in Pittsburg in which the researchers combined 3 technologies: off-the-shelf face recognizer called (Acquired by Google), cloud computing and the information that is publicly available in social networking sites.
Using photographs of students Acquisti?s team identified people on a popular online dating site where the members used pseudonyms to protect their privacy. In the second, students walking on campus were identified using their profile pictures on facebook. In the third experiment the research team was able to predict personal interests and even the Social Security numbers of students, simply with a photo of their face. The professor is quoted saying, ?we call it the democratization of surveillance?
93 students volunteered to take pictures of themselves on web-cams attached to laptops. Their photos were then uploaded to a cloud computer, and compared to a database of 261,262 publicly available photos downloaded from Carnegie Mellon student?s profiles on Facebook. In less than 3 seconds the face-recognition tool was able to yield 10 possible matches. The students confirmed that for more than 30% of the time their faces were among the top results.
In 1987 the Social Security Administration started assigning numbers in a way that made it easy to predict them by knowing some one?s birth-date. Taking this into consideration the researchers tried to predict the Social Security Numbers of their participants. About 27% of the time, taken from information found on Facebook, of those identified, he could predict first 5 digits of their Social Security numbers correctly with just 4 attempts.
The last 4 digits of the number were also predictable. In 2009, a paper by the Professor Acquisti showed that the Social Security number of individuals born after 1988 could be predicted within 1000 attempts for up to 10% of the people. Subsequently, in June the Social Security agency launched a new “randomized” numbering system, making such predictions more difficult for future generations.
Face recognition technology has taken a steep climb recently. The number of photos that have been incorrectly rejected by face recognition tools has fallen to a mere 0.2%? in 2010 from 79% in 1993 according to a study by the National Institute of Standards and Technology. What fueled this step forward is the amount of information available online through various social networking sites. Users upload their photos which initially were only viewable in personal belongings such as driving licenses.
Facebook released facial recognition service world wide in July. This allowed people to automatically identify photos of their friends. Users can change their privacy settings if they do not want to be automatically identified.
At D: All Things Digital conference in June Eric Schmidt (Google?s chairman) expressed his concern about Facebook. It is “the first generally available way of disambiguating identity,” he said. ?A Facebook spokesman said that the users of the social networking site have the right to choose if they upload a profile picture, it?s content and if they want to delete it. Therefore, it is not always the case that an individual?s profile would contain the picture of their face. The website has 750 million users, for this to be affective the users must be aware of this option.
According to a Google spokesman, Google does not intend to release the facial recognition software with out strong privacy protection in place. At the D conference Schmidt said that Google h ad withdrawn a facial recognition service for mobile phones that it considered to be too intrusive.
The Carnegie Mellon research team extended the experiment by creating a mobile app that takes pictures of an individual and overlays on it prediction of person?s name and social security number. It was stated that they do not choose to release the application. The results indicate that all of us may be able to be recognized on the streets by anyone with internet connection and bring into question just to what extent ones privacy can be protected in this fast growing world.
Sources: The Wall Street Journal, Carnegie Mellon University, P.C.Mag