Software To Identify Islamic State Propaganda Developed By Georgia State Researcher, Collaborators

ATLANTA—Software that can be used to identify Islamic State or Da’esh propaganda videos by analyzing the music embedded in their soundtracks has been developed by a Georgia State University terrorism expert and his collaborators.

“We realized that this method could be a potentially important contribution to efforts to flag and identify terrorist content online, based on the audio contained in it,” said Anthony F. Lemieux, director of the Global Studies Institute, and professor of global studies and communication at Georgia State.

In the study, Lemieux, who is also a member of Georgia State’s Transcultural Conflict and Violence Initiative, worked with Jonathan Pieslak, a professor of music at City College of New York and Brian Pieslak, a partner of Web and mobile design firm Sedulen, to create an automatic content recognition (ACR) tool, similar to the popular music app Shazam. The software they created, named Marvin, allowed the researchers to single out Da’esh-produced Islamic songs (or anashid) and Koranic recitation appearing in Da’esh video.

“Specifically, if a video has music embedded in it that is a product of the Islamic State’s music production wing of their media outfit, then it would really be something that should be screened further and most likely taken offline,” he said. “Our approach adds an element of identifying and flagging content that should be examined more carefully, and we’ve shared these ideas with entities, including Facebook and the Home Office in the U.K., where identifying and removing terrorist content has been a huge priority.”

The researchers assembled a library of more than 1,200 jihadi and jihad-themed audio products and categorized Da’esh- and Al Qaeda–specific content in the library. In the interdisciplinary collaboration involving a musicologist, a software architect and a social psychologist, the team developed the Marvin application to enable searches via a Cloud service that provides audio fingerprinting by comparing new audio clips to samples in the library. Because the database is exclusively made up of content produced by terrorist groups, the odds of mistakenly tagging legitimate Koranic recitation are small, Lemieux said.

“ACR technology represents a useful avenue for recognizing specific jihadi media content posted online,” Lemieux said. “The recognition of this highly specific audio could signal a terrorist video in addition to visual or video recognition tools. If successfully applied at scale, this kind if technology could significantly help to curb the dissemination of online terrorist propaganda.”

The research team included several Georgia State Honors College undergraduate students. The study was supported by the Minerva Research Initiative of the U.S. Department of Defense, through the Air Force Office of Scientific Research as part of the Mobilizing Media research program.

The article, “Trends of Anashid Usage in Da‘esh Video Messaging and Implications for Identifying Terrorist Audio and Video,” was published the academic journal Studies in Conflict & Terrorism.

Featured Researcher