Forensic stylometry has been used to discover the writings of authors ranging from James Madison to J.K. Rowling. At the same time, this forensic technique has been at the center of some high profile blunders.
Welcome to the Fourandsix blog, where you’ll find tips on image forensics techniques and commentary on issues relevant to photo tampering and the responsible use of imaging tools.
In 2010 the television commentator Glenn Beck held a rally at the Lincoln Memorial. Rep. Michele Bachmann claimed over 1,000,000 were in attendance. Beck claimed a crowd of 500,000, and NBC News counted 300,000 in attendance. At the same time, the firm DDIS, hired by CBS News, arrived at a head count of 87,000. Politics often gets in the way of counting crowd size. Recent advances in image analysis and aerial photography, however, have begun to improve the reliability of such measurements.
Over the past few years, Twitter has become one of the leading disseminators of information during breaking news events. From street protests in Egypt, to helicopter crashes in London, and hurricane damage in New York, everyday citizens and their smartphones are often there much earlier than professional journalists, and in far greater numbers. Unlike professional journalists, however, these Twitter users are often anonymous, and sometimes little can be gleaned about their reliability, methods, or motivations. As a result, it shouldn’t be too surprising that there’s lots of erroneous information mixed in with the facts. A variety of professionals and organizations—including Fourandsix—have been hard at work on methods to help separate the real from the fake. Recently, some new research that proposes a method to isolate fake photos based on characteristics of the accompanying tweets has received some attention as a potentially promising idea.
Why do we pay closer attention to a bear than to a rock? Because, obviously, a bear poses more of a threat to us than a rock (unless someone is throwing the rock at us). In fact, vision scientists have shown that images of people, animals, or faces in general, grab our attention faster than inanimate objects like a rock. One might imagine, however, that a doll’s face, for example, will inspire a different response than a living human or animal face. A team of Dartmouth scientists (Dr. Christine Looser and Prof. Thalia Wheatley) now provide some insight into how we process images of inanimate and animate faces. This interesting research has implications for the task of distinguishing between computer generated and photographic images.
The same researchers that brought us a technique for determining the location that a photo was taken now bring us a technique for determining the approximate date that a photo was taken. Using statistical measurements from a single color image, this technique automatically determines the decade in which a photo was likely taken.