Researchers have developed a more comprehensive method of authenticating smartphone photos.
Duke computer scientists recently designed and developed YouProve—a system of validating the authenticity of images and sound clips captured with smartphones. The technology detects changes and creates a report documenting alterations. YouProve was created by a team of students led by Landon Cox, assistant professor of computer science and electrical and computer engineering, who worked with researchers at Microsoft Research and Technicolor Research.
“With citizen journalism, there are great opportunities to report on a lot of interesting events where there don’t happen to be [professional] journalists,” said Peter Gilbert, a second-year graduate student in computer science. “The authenticity for the integrity for these kind of reports from smartphones is already becoming a concern.”
Media recorded on smartphones is integral in describing events as they happen, Gilbert said. He added that news outlets are continuously trying to find new ways to bring news to the masses, demonstrating a need to ensure that amateur media is accurate and unaltered.
Increasingly, news organizations are relying on smartphone submissions to show a first person perspective of events. As such, organizations also run the risk of using doctored images, Gilbert said.
“YouProve uses a combination of trusted hardware and operating system extensions to monitor how applications modify media captured on a mobile phone,” Cox wrote in a Nov. 15 email. He added that YouProve is not a web application because it requires access to a device’s hardware to monitor media files and manipulations in order to provide a text file report on the findings.
The technology correctly identified altered portions of images and audio clips with 99 percent accuracy, producing the analyses in less than 70 seconds.
The team considered several other factors while designing the technology, including possible infringements on people’s privacy.
“Our motivation was—how do you preserve the integrity of an image while not violating [a person’s] privacy?” said Kyungmin Lee, Trinity ’11.
Gilbert noted that previous attempts at similar programs provided very general information, such as whether or not an image had been edited in any sort of way.
“Sometimes you have to modify an image somehow,” Gilbert said. “Due to bandwidth constraints, you are probably not going to be able to upload this full, uncompressed... photo to a news website.”
He added that there are many legitimate modifications to photos—including cropping, resizing or scaling a photo down. One of the other major concerns is the privacy of not only the sharer of the photo but of the people in it.
“A photographer might want to crop out or blur the faces of the people in his [smartphone] photo,” Gilbert said. “Sometimes these images are taken in locations where the safety of the subjects could be compromised if their identity is revealed in the media.”
YouProve takes such factors into consideration and allows contributors to have the freedom to crop and blur images, which will then be noted in the report generated by the program, Gilbert said. The report identifies finer areas of the photo that are changed in some way.
YouProve is currently a system of extensions only for Android phones, because Android is available under free and open source software, Gilbert said.
Researchers hope the methods of YouProve are general enough to be applied to the Apple iPhone in the future, but the source code for the iPhone’s operating system is needed first.
YouProve relies on a special hardware component—Trusted Platform Module—that is not available on all smartphones today, but is on a majority of laptops. As soon as more TPM-inclusive smartphones are on the market, Gilbert and Cox noted that they would like to expand YouProve to additional platforms.
Get The Chronicle straight to your inbox
Signup for our weekly newsletter. Cancel at any time.