Thursday, August 18, 2011

Facebook's facial recognition system, why it's scary

Oh, iPhoto, your facial misrecognition antics ...Image by Chris Devers via FlickrI'm going to be posting several posts on this information. Somthing I feel could possibly bring some very dangerous outcomes to many people who don't pay attention to this.
Not out of fear, but so that you the general person may be informed on what is really happening with this technology. Be sure to check out all the links and more will be posted on the home page at the top as well.
(CBS) - Remember last weekend, when you were at that party and friends were snapping photographs to post on Facebook? What if that seemingly innocent act could lead to identity theft or real-life stalking? Fear!
In a recent study done at Heinz College, Carnegie Mellon University (CMU) called "Faces of Facebook: Privacy in the Age of Augmented Reality," Alessandro Acquisti, Ralph Gross and Fred Stutzman reported on the "consequences and implications of the convergence of three technologies: face recognition, cloud computing, and online social networks."
Presented on Thursday at the Black Hat Technical Security Conference, the study argued that with current technology, we could be "re-identified" and have our social security numbers stolen. The three experiments discussed are eye-opening.
Facial recognition and re-identification
The first experiment took pictures from a popular dating site and cross referenced them with Facebook profile pictures. Dating site members typically maintain their anonymity by using pseudonyms. With the results of the experiment being "statistically significant," these people could easily be identified with the software.Thus their identities outed across both channels.
The second experiment was similar to the first, except this time, the researchers took real-life photos and used facial recognition software to re-identify students at a select college campus. They were able to identify one-third of the subjects in their experiment.
Augmented reality
In their final experiment, the researchers used the phrase "augmented reality" to describe "the merging of online and offline data that new technologies make possible."
The study's goal was to show "that it is possible to start from an anonymous face in the street, and end up with very sensitive information about that person." The results were "made possible by the convergence of face recognition, social networks, data mining and cloud computing - that [is referred] to as augmented reality."
(Credit: AP Photo/Gerry Broome)
For example, once a photo is snapped on the street, facial recognition software can be run to identify a person online. Some data mining could possibly uncover the city and year the subject was born. A little knowledge of how social security numbers get assigned will enable identity thieves to narrow down the range of numbers to work with.
See more here.

The Dangers of Using High Tech Facial Recognition Software to Catch Criminals
In the wake of rioting in London, England, the British police are turning to CCTV footage in an attempt to capture the ringleaders involved in the rioting and troublemaking. With the advances of high-tech facial recognition software used in airports to identify people at immigration, they say it should be possible to match up faces with know criminals and troublemakers. But will it be that easy?
There are attempts being used, albeit private sector attempts, to try and use this software and match people with that of criminal records and known photographs of perpetrators, but it’s not altogether that simple really, because in order to do facial recognition you already need a database of confirmed photos, with names attached to them. You can’t simply look at a photograph and work out who it is, and you must of course have something to compare the CCTV footage with.
The other issue is the quality of the photograph you’re trying to compare from. There is a huge difference between a photo of someone staring straight at a camera with a plain background, and another which is slightly misty and possibly from a significant distance away, plus the fact that many rioters in London wore masks to cover parts of their faces. It actually turns out to be very difficult indeed to match these kinds of photos, and as of yet the technology seen on sci-fi movies where you can enhance pictures to extraordinary degrees is not available. Unfortunately real life isn’t quite like that, so the dodgy quality CCTV footage the police have may not be entirely reliable and open to interpretation. Certainly in a court of law, a fuzzy picture may not stand up on it’s own without other evidence to support it, such as eyewitnesses. So if we have to rely on eyewitnesses to get a conviction, we’re really back to square one.
Again see more here.
What is Facial Recognition?
A general statement of the face recognition problem (in computer vision) can be formulated as follows: Given still or video images of a scene, identify or verify one or more persons in the scene using a stored database of faces.
Research directions (according to Face Recognition Vendor Test - FRVT 2002):
  • Recognition from outdoor facial images.
  • Recognition from non-frontal facial images.
  • Recognition at low false accept/alarm rates.
  • Understanding why males are easier to recognize than females.
  • Greater understanding of the effects of demographic factors on performance.
  • Development of better statistical methods for understanding performance.
  • Develop improved models for predicting identification performance on very large galleries.
  • Effect of algorithm and system training on covariate performance.
  • Integration of morphable models into face recognition performance.
  • Understanding the video sequences in FRVT 2002 did not improve performance.
Since the former Face Recognition Homepage of Peter Kruizinga is no longer active and we believe that face recognition is an important research area, we wanted to give prospective researchers a place to start on the Internet. Our Face Recognition Homepage aims to provide scientists with the relevant information in the area of face recognition. This page is intended to be an information pool for the face recognition community. Its goal is to provide an entry point for novices as well as a centralized information resource to concentrate face recognition and related scientific efforts.
The materials currently published on the web-site are what we believe every researcher in this area should read. Web-site is designed to be a portal, so you will notice numerous links to other research groups, links to relevant journals, books, databases etc. Another issue also worth mentioning is that this is a non-commercial, purely academic web-site.
On this Face Recognition Homepage you can find:
  1. some general information about face recognition problems;
  2. information about all changes on this Face Recognition Homepage;
  3. links to research groups that work in the area of face recognition;
  4. link to Face Recognition Research Community newsgroup (established in January 2007), where researchers can post questions and exchange ideas;
  5. interesting papers that deal with the face recognition (general papers, standards, cognitive vision / psychology / neuroscience papers, highly cited papers, published items vs. citations);
  6. new papers with most recent advances (the state-of-the-art) in face recognition;
  7. face databases often used by researchers;
  8. face recognition algorithms (image-based and video based) with a few most representative papers for each algorithm;
  9. source codes contributed by users (Matlab, C/C++, Java);
  10. list of conferences, where face recognition is the conference topic;
  11. links to high impact factor journals which scope covers face recognition; information about some journal special issues; books on face recognition; other related books;
  12. links to vendors developing face recognition technology;
  13. other related links;
  14. contact information.
More coming very soon. If you don't see a link here yet. Be sure to check out the home page to see more.
Enhanced by Zemanta



Related Posts with Thumbnails

wibiya widget