The defalsif-AI project: protecting critical infrastructures against disinformation and fake news

In this paper, we describe the concept and ongoing work of the project defalsif-AI, which addresses the protection of critical infrastructures against disinformation and fake news. Defalsif-AI deals particularly with the protection of the main democratic processes and the public trust in democracy and its institutions against engineered social media attacks, which, for example, attempt to manipulate the electoral process. Federal ministries and media institutions require new methods and tools to evaluate the ever increasing amount of digital media in terms of identification, verification, and correction of sources. Based on these requirements, the project focuses on research on audio-visual media forensics, text analysis, and multimodal fusion with the support of artificial intelligence (AI) and machine learning methods. One main focus of this research is to make the results more comprehensible and interpretable for non-experts in the forensic/technical field. The primary project outcome is a proof of concept of a multimodal detection platform, which can operate with a variety of sources, including the surface web and social media. Additional research carried out within the project focuses on providing and generating multimodal data necessary to train and test machine learning models. Finally, an analysis and assessment concerning the law and social science are carried out as well.