Project TGuard

Hybrid threats such as disinformation campaigns and the targeted dissemination of false reports are among the major – and often underestimated – dangers facing society. Through the "TGuard" research project, a team at FH Salzburg is shedding light on how disinformation spreads across digital networks.

The fact that attacks are becoming increasingly frequent and ‘smarter’ also applies to the media and communications sector. Whereas in the past it was possible to expose manipulation with a little attention and common sense quickly, artificial intelligence now makes it much more difficult to uncover fake news.

As part of the interdisciplinary research project TGuard, Clemens Havas and his team from the Department of Creative Technologies are working with partner institutions to develop a demonstrator application that illustrates how artificial content is generated, how social media bots work and how disinformation spreads across digital networks. In a specially created simulation environment, workshop participants experience for themselves just how easy or difficult it is to spread and detect disinformation.

Targeted disinformation, or the creation of information bubbles within which individual groups operate, has long since ceased to be an individual problem and has become a societal one.

Examples:

“This video shows the network graph of the TGuard demonstrator. Each point, also known as a node, represents a user, and each line, also known as an edge, represents a ‘follow’ connection,” explains Clemens Havas.

At the start, the network is barely connected; there are many individual users on the network who have not yet established any fixed connections. Over time, users get to know one another through chance interactions (for example, they might see a post in the global feed or a comment under another post) and can follow one another if they share similar views. 

As is common on social media networks, some users have a disproportionately large number of followers. In the real world, these would be, for example, public figures or official accounts belonging to companies, governments and the like. In this network, preferences arise purely from the agents' behaviour but lead to a similar result. 

In the final part of the video, one can see just how great the influence of individual nodes in the network is. User Teun de Vries has many edges, is therefore well-connected and wields a great deal of influence within the network. The users Alexei Ivanov and Luc Van den Broeck, on the other hand, both have relatively few connections and therefore do not have a particularly significant impact on the network.

Quiz

Test your AI knowledge with the game ‘Which face is real?’, developed by Jevin West and Carl Bergstrom at the University of Washington.

Still unsure? The quizzes and training courses on Saferinternet.at can help.

Project partners

In addition to the FH Salzburg, the project involves the Federal Ministry of Defence, AIT, ÖIAT, AIES and neke-neke GmbH. Funded under the K-PASS cyber security research funding scheme of the Federal Ministry of Finance.