News | Aug 1, 2019

Air Force software using Markov logic finds propaganda, disinformation, and malware on Twitter

"The power of social media is undeniable...hackers can post false information"

August 1, 2019, update: An additional, related patent application from the U.S. Air Force Research Laboratory was revealed today. The new filing is titled “Method of Cyberthreat Detection by Learning First-Order Rules on Large-Scale Social Media.” The document details the inventors’ improved system for automatically detecting malicious accounts on Twitter. Download the new patent application below. Contact Sean Patten at spatten@montana.edu or 406-994-7721 for more information.


July 9, 2019, update: The U.S. Patent and Trademark Office issued U.S. Patent 10,348,752 to the Air Force Research Laboratory today for the invention described in this story–a link to the patent is provided below.


Dec. 20, 2018 – Sniffing out disinformation and malware on Twitter can be accomplished through random variable probability statistics, according to the Air Force Research Laboratory’s Information Directorate in Rome, New York.

Computer scientists at the Air Force lab used Markov logic networks, named for Russian mathematician Andrey Markov, to power an application that screens Twitter’s firehose of information for propaganda and cyber threats that threaten the nation’s economy and security.

Last month, the government published an Air Force patent application on the technology, making it available for license to private-sector partners who would like to offer it as a new or improved product.

Because “the nature of cyber attacks on social media is quite complex” the Air Force’s system develops its own “knowledge base on social media posts” that can model “various kinds of possible attacks by adversaries on Twitter using first-order logic within a single, unified framework” and then identifies potential threats using probabilistic inference.

Variables in the formulas used to create the knowledge base include friendship relations, trending hashtags, retweet counts, verified user followers, total follower counts, hyperlinks, user tags, and the number of likes of non-follower tweets.

“Note that when a user’s tweet is liked by someone, then a notification is sent to the user. Thus, a suspicious user can draw the attention of other users to himself/herself by randomly liking their tweets,” the patent states.

Air Force inventions, including software, are available to private industry for use in new or improved products and services through a business-friendly process called technology transfer, which typically produces a patent license agreement.

TechLink, the Department of Defense’s national partnership intermediary for technology transfer, and the Griffiss Institute, a nonprofit foundation located next to the laboratory in Rome, have helped hundreds of companies license software from the Air Force.

Businesses and entrepreneurs can receive licensing assistance from the two organizations at no charge, said Sean Patten, software licensing lead at TechLink.

“The Air Force’s lab in Rome has a long history of impactful technologies and technology transfer success,” Patten said. “This is an opportunity for businesses that want to provide software as a service to public or private customers.”


Contact Sean Patten, spatten@montana.edu or 406-994-7721.