The FERMI consortium held its 3rd in-person meeting in Helsinki, Finland on 27-28 September 2023. The meeting was hosted by the Finnish Ministry of the Interior (FMI), one of the law-enforcement agencies (LEAs) amongst the FERMI partners.
The agenda largely focused on the project’s technical work, which is primarily aimed at producing a platform consisting of different tools to rein in the crime-related ramifications of spreading disinformation on social media. For the first time a preliminary version of the FERMI platform was demonstrated; a schedule that was well in line with FERMI’s Grant Agreement (GA) that required the platform to be available in the form of a Minimum Viable Product (MVP) by the end of September 2023.
Updates on the Technical Development
The MVP version of the platform includes numerous components. In accordance with FERMI’s outline, they are all aimed at supporting LEAs in the fight against disinformation-induced crime. The contributions they make to this effort, however, vary greatly and are meant to supplement one another in the sense of supporting different countermeasures. The development of each of the components below is addressed in separate tasks in the more technical Work Packages (WP) of the project (WP3 and WP4):
- Dynamic Flows Modeler – Crimes Impact Predictor: predicts the crime landscape in the aftermath of disinformation campaigns taking place on social media
- Disinformation sources, spread and impact analyser: identifies the further spread of disinformation messages by showing a graph of all reposts/ retweets and distinguishes between messages that most likely have been shared by humans and those that – presumably – have been posted/ tweeted by bots
- Community Resilience Management Modeler: assesses the resilience of communities to the spread of disinformation
- Swarm Learning Module: designed to allow multiple LEAs to collaboratively train machine learning models without sharing raw data centrally, which might greatly enhance the predictions made by the platform
- Behaviour Profiler & Socioeconomic Analyser: grasps the likelihood and severity of the costs of politically motivated crimes such as those caused by the above-mentioned disinformation campaigns
- Sentiment Analysis module: analyses the sentiments of the social media messages at stake and the reposts/ retweets alike
- FERMI Socioeconomic Disinformation Watch: merges the Behaviour Profiler & Socioeconomic Analyser and the Community Resilience Management Modeler to rank the threat posed by disinformation campaigns
The development of all of these components was deemed well on track. The outstanding issues mostly concern data acquisition and, to a lesser extent, a few integration challenges that require some further technical effort. The possible impact of data acquisition issues is marginal at worst in the vast majority of cases and integration issues seem unlikely to pose insurmountable challenges to the project.
Moreover, ethics requirements were reported to have been addressed, communication and dissemination activities were found to be well underway. The project has clearly exceeded the expectations for the first 12 months as derived from the GA’s overall demands. The pilots and validation efforts are currently being planned.
The FERMI partners also elaborated on the next steps. The above-mentioned components will be further fine-tuned and integrated, an outline for conducting the pilots will be drafted, exploitation efforts will be embarked on more comprehensively (in line with the GA’s schedule that requires the exploitation-focused task to start in October 2023), and communication and dissemination activities will continue to build on the ground that has been laid before. One of the key objectives in the near- to mid-term future is to present a fully integrated version of the FERMI platform by March 2024.