*Interesting line of work.
They eat the blue pill and they believe whatever they wanna believe
BY CHRISTYE SISSON5 MINUTE READ
Lots of people—including Congress—are worried about fake videos and imagery distorting the truth, purporting to show people saying and doing things they never said or did.
I’m part of a larger U.S. government project that is working on developing ways to detect images and videos that have been manipulated. My team’s work, though, is to play the role of the bad guy. We develop increasingly devious, and convincing, ways to generate fakes—in hopes of giving other researchers a good challenge when they’re testing their detection methods.
For the past three years, we’ve been having a bit of fun dreaming up new ways to try to change the meaning of images and video. We’ve created some scenarios ourselves, but we’ve also had plenty of inspiration from current events and circumstances of actual bad guys trying to twist public opinion.
I’m proud of the work we’ve done, and hope it will help people keep track of the truth in a media-flooded world. But we’ve found that a key element of the battle between truth and propaganda has nothing to do with technology. It has to do with how people are much more likely to accept something if it confirms their beliefs.
FINDING, AND PUSHING, TECHNICAL BOUNDARIES
When we make our fakes, we start by collecting original, undoctored images and videos. Those not only offer raw material for us to manipulate the images but also include the data stored in authentic media files—sort of like a technical fingerprint that accompanies every piece of media that describes how and when it was taken, and with what tools.
That information helps us craft fakes that look and act as much as possible like real material, in both visual evidence and digital artifacts. It’s an ever-changing challenge, as new cameras go on the market and as researchers develop new techniques for digital forensic analysis.
What we create are then sent to other research partners in the larger effort, to see if they can tell what we’ve done and how we’ve done it. Their job is not just to determine whether it’s authentic or fake—but also, if possible, to explain how the fakes were made. Then we compare the results to what we actually did, and everyone learns; we learn how to make better fakes, and they learn to detect them.
