When a picture of John Kerry and Jane Fonda speaking together at an anti-war rally appears on the Internet, who is to say that the event never took place -- that in reality the two speakers never shared a platform?
Thanks to the pioneering research efforts of Hany Farid, a computer science professor at Dartmouth, analysts are gaining access to technology that allows them to identify manipulated digital images such as the Kerry/Fonda collage.
Just yesterday, Farid sent code he has developed for authenticating digital media to the forensics department of the FBI, in the hopes of helping them crack down on digital forgery -- a problem that has grown in the past few years thanks to technological advances in Photoshop and the Internet.
According to Farid, there are a number of different tools that one might use to manipulate an image. Farid and his team of researchers, graduate students Alin Popescue and Siwei Lyu, investigate these tools one at a time.
They begin with an image, digitally remaster it in a specific, isolated way, and in the process, record the mathematical formula they used to do so. In analyzing what mathematical patterns they disturbed, they are then able to retrace their steps and develop a computer program that can detect these disturbances.
To date, Farid and his team have written six such programs, and he says many more are in the works.
Farid plans to go to Boston next week to present his findings to a Massachusetts associate district attorney general in the hopes of putting his research to work in the courtroom.
There are currently no laws regulating digitally-remastered evidence. As the statute now stands, an image is admissible if it is recognizable by sight, regardless of the possibility of its being a forgery.
Farid has been asked to consult on several cases and has done work with federal and state prosecutors. While his trained eye is capable of noticing cognitive inconsistencies, he is wary of testifying that an image has been doctored unless he has the mathematical data to back up his findings.
Among the quantative criteria one looks for in a forgery are different levels of graininess -- called noise characteristics in the field -- often resulting from splicing together pictures taken by two different cameras.
Farid has spent the past three years studying digital tampering and forensics.
He found inspiration for his research while "being a nerd and poking around the library at MIT," where he earned a postdoctorate in brain and cognitive science.
After reading that courts do not have digital evidence standards, he was determined to discover whether it was possible to use computers to authenticate such evidence.
As other scientists undertake this research in the next few years, Farid said he will most likely move into other areas of study. He is concurrently studying the digital reconstructions of Egyptian tombs and working with another graduate student on image-guided neurosurgery.
Funding for Farid's research comes from the National Institute of Justice and the Dartmouth College Cybersecurity Research Group.
Because all of his information is Dartmouth College property, he is obliged to openly share it with federal institutions such as the FBI.
When asked if he would have liked a patent and more recognition, Farid smiled and shook his head.
"I am an academic in the business of doing science and educating people; my interest is to do good work and make it available to people who need it," Farid said, leaning casually back in his chair. "I have a pretty cool job if you ask me."
