 |
 |
 |
 |
 |
 |
 |
 |
 |
Hany Farid, director of Dartmouth College's Image Science Laboratory, is one of the pioneers of digital forensics, a computer-science subfield that specializes in determining if digital images, audio, video, or documents have been manipulated. After obtaining his Ph.D. in computer science from the University of Pennsylvania in 1997, Farid held a two-year post-doctoral position in Brain and Cognitive Sciences at the Massachusetts Institute of Technology. He then joined the Dartmouth faculty in 1999, where he serves as the David T. McLaughlin Distinguished Professor of Computer Science. Farid is currently working with law-enforcement officials to develop advanced software to detect if photographs have been altered. |
On July 1, 2008, Hany Farid answered selected viewer questions about photo fakeries, the promise of digital means to detect manipulated images, and more. Please note we are no longer accepting questions, but see "Fake or Real?" and our Links & Books section for more information. Q: Is your photo on this page a composite? The white-spot locations are different for both eyes—your left eye has more spots. Also, each half of your face, when viewed separately, has its own personality—the left side is perky and cocky-looking, while the right side seems tired and the smile looks forced. A: Stephen, you have a good eye. The photo is not a composite, but it has been altered to enhance the contrast and color. You are right that my right eye shows only the reflection of one light, while my left eye shows the reflection of two lights. This is because the angle between my right eye and the second light is such that the reflection landed outside of my iris. The human face does generally have asymmetries, which is what you are noticing (and apparently some days my personality has asymmetries as well). Q: Is it possible to fake an image so well that it can trick digital-forensics software? There must be people who are working as hard to defeat forensics detectives as you are trying to tease out the forgeries. A: Scott, although not easy, it is definitely possible to create a fake good enough to fool all modern digital forensic techniques. I have created such fakes. As with the spam/anti-spam and virus/anti-virus game, the forensics/Photoshop game will continue to evolve with both sides becoming more and more sophisticated as the years go on. Q: How do you think issues of digital fakery will impact the judicial and legislative process in the future? What advice do you have for those of us working in the law-enforcement sector? A: Lee, the impact of digital fakery is already affecting the judicial and legislative process. For example, the U.S. Supreme Court recently ruled that virtual (computer-generated) child pornography is protected under the First Amendment. In so doing, the Court's decision has made it difficult to prosecute violations of the child pornography laws, since it is easy to claim that images are not real, and difficult to prove that they are. Similarly, it is easy for anyone to simply claim that an audio recording, image, signed document, or video has been altered, leaving the court to contend with complex questions of authenticity. As technology continues to advance, and as digital imaging becomes more ubiquitous, these issues will only become more widespread and more complicated. As a result, it is important that the law-enforcement sector keep up with the fast pace of technology. And it is equally important that the field of digital forensics evolve to help the courts and other sectors contend with these issues. Q: I work for a police-department photo lab, where we are very interested in obtaining your software. When will it be available? Can anyone get a copy? A: Joe, we are in the early stages of commercializing our image forensic software and hope to have a version ready for law-enforcement agencies some time next year. Q: Could your software be used to uncover hidden things in videos, such as the movement of lips behind a beard? A: Al, our work does not directly address this type of question, but there are techniques in image and video enhancement that might be applicable. Q: Can your software provide any information on more subtle tweaks done on real photos, such as brightening, light airbrushing, or age progression? A: Ethan, given the original resolution image, some of our techniques can determine if an image has been manipulated in any way from the time of recording. With a reduced resolution or heavily compressed image, other techniques are not as effective at determining more subtle alterations such as brightening or color adjustments. Q: What is the minimum resolution or the smallest image size that your system can work with? A: Nate, in general, the higher the resolution of the image, the easier it is to apply many image forensic tools. While there is no absolute lower limit, images smaller than 200 x 200 pixels begin to pose significant challenges for most forensic techniques. Q: Have you looked at depth of field to identify fakes? What is the likelihood that two different photos, presumably taken with two different cameras, will have precisely the same depth of field? A: Deborah, this is an interesting question. The aperture size is the primary determinant of depth of focus. (A large aperture yields a narrow depth of focus, while a small aperture yields a large depth of focus.) It is quite possible that photos taken with different cameras yield the same depth of focus—particularly with point-and-shoot cameras, where the range of aperture sizes is limited. The depth of focus can, however, be a useful visual clue to determine the conditions under which an image was taken. For example, a large depth of field, where nearby and distant objects are equally in focus, implies that the image was photographed with a small aperture. A small aperture implies a small amount of light entering the camera, which means that a relatively long exposure time was needed to provide enough light to properly expose the image. A long exposure time often leads to motion blur (objects that are in motion being blurred in the image). Inconsistencies in these visual clues should raise questions of authenticity. Q: How can someone with a limited knowledge of photo-editing software determine if a photograph is fake or real? Are there any "tricks" the average Joe should know? A: Joe, there are some visual clues that can be useful in looking for fakes. (Go to the Fake or Real? audio feature for some hints.) Differences in lighting, differences in focus, differences in image quality, can each be useful in spotting a fake (see also Deborah Haddow's question above). However, care must be taken, as sometimes legitimate photos may, at first glance, appear to be fakes (see, for example, the response to the first question). Q: In your opinion, what is the best photo manipulation software? A: Adobe Photoshop is one of the most sophisticated and most popular photo editing software packages. We routinely use Photoshop in my lab to manipulate photos that we then use to test our forensic analysis tools. Q: There has been a lot of controversy over the decades on the "backyard photographs" of Lee Harvey Oswald holding a rifle and some Communist Party newspapers. Do you know if these images are authentic? If you haven't looked at them yet, do you plan to? A: Jim, I have looked at these photos. While there are some peculiar lighting artifacts, they do not necessarily point to photo manipulation. The quality of the photos I have seen is too poor to permit a complete analysis. Q: Did Robert Peary really reach the North Pole? There is a famous photograph that purports to show Peary and his companions very near 90 degrees north. But some scholars have called into question whether the direction of the lighting is consistent, astronomically. You simply must bring your expertise to bear on this fascinating historical controversy! A: George, I have not looked at these photos. We have, however, analyzed the original moon landing images and found that the lighting in all of the photos is consistent, including consistent with the expected position of the sun given the landing location and the date. Q: Do the Associated Press or major newspapers like the Los Angeles Times ever check to see if the photos they publish or send through the wire have been manipulated? A: Don, the Associated Press processes over 10,000 photos each day. While I imagine that a photo editor will look at some of the photos prior to publication, it is simply impossible to manually analyze all of these photos. We are working on streamlining our forensic software so that media outlets can process a large number of photos in a relatively short amount of time. Q: Why would a magazine or anyone else fake a picture? What do you think their motives usually are? A: Devin, I think that some magazines are simply trying to create a picture that fits with their story. For example, Newsweek spliced a picture of Martha Stewart's head onto the body of a slim model to accompany the headline "After prison she's thinner, wealthier & ready for prime time". And others just want to sell magazines by splashing a sensational picture onto their magazine's cover ("Brad & Angelina, caught together!"). Q: What college courses would you recommend for someone interested in someday working in the digital forensics field? A: Josh, I would recommend a combination of computer science, applied mathematics, and electrical engineering. Not only will these fields of study provide you with the necessary tools for a career in digital forensics, but they will also provide a broad and generally useful base for future study in many other disciplines. This is particularly important since technology shifts quickly, and you want to create a strong but flexible foundation that allows you to change with technology. |
||||||||||
|
|||||||||||
| |||||||||||
© | Created June 2008 |
|||||||||||
