One of the main goals of the Joint Action “Research – Create – Innovate”, is to strengthen research that involves innovation and can create entrepreneurship. The desired results of the action include the strengthening of the competitiveness, productivity and extroversion of the enterprises and the transition to quality innovative entrepreneurship and the increase of the domestic added value. In this context and with the proposed project, the participating companies aim to develop tools for:
a) the identification of sources / nodes involved in online propaganda activities and their classification into different categories (false news sources, retransmission nodes, news paraphrasing / distortion nodes, etc.);
b) the detection of organized astroturfing acts as they are shaped in social media.
Utilize the know-how, content and tools of both PALO and Qix and develop new competitive services that will enrich the existing range of services for products and businesses, and upgrade their existing business product analysis corporate image on social media. Existing products and services have services developed jointly by the two companies under the ICT4GROWTH action while individual tools have been enhanced by other Competitiveness OP actions.
c) to offer unique technological solutions to problems of global interest such as fake news and a specialized approach to fake news dissemination strategy through structured networks of influence (as in the case of astroturfing) and is expected to create opportunities for expansion at national and regional level and collaborations with companies abroad.
This project is part of the theme “8.7 Emerging Technologies: Cybersecurity and Internet Trust” and specifically the priority & quot; 8.7.2 Reliability and quality of information and web profiles “because:
● It will develop tools for checking the reliability of news transmission and retransmission sources (nodes) on social networks which will evaluate the information circulated and transmitted by the sources, the transmission and retransmission rate, the information dissemination paths in which they participate.
● Will train a single forecasting model that will evaluate news releases, reposts and commentary and assess the veracity of the news based on its dissemination pattern. It will train graphic-synergistic neural networks at the first level to model the dissemination of news in source graphs and network profiles and then use reinforcing learning to decide on each degree of validity based on how it was disseminated.
● For this purpose, it will examine social media sources that produce content (news sources, blogs, etc.) of different validity and credibility by topic category, social network nodes (users) that reproduce this content or comment on the produced content in different in and will use positive and negative training samples. It will then offer a tool that will be constantly trained each time a news item is flagged as false.
● Attempt to capture in detail the propaganda networks as an extension of the simple flag lists (or black lists) formed behind a deliberately false news or news group that systematically targets an entity and develops a method that locates the sources and their transponders of a propaganda as well as other news that circulate on the same networks.
1 Deutche Welle, “G7, Facebook, Google, Twitter agree on plan to counter Islamist terror”. Available at https://www.dw.com/en/g7-facebook-google-twitter-agree-on-plan-to-counter-islamist-terror/a-41055080. Published on the 20/10/2017.
2 Wall Street Journal, “Tech Firms Face Prospect of Fines in Europe Over Terror Propaganda”. Available at: https://www.wsj.com/articles/eu-backs-fines-for-tech-firms-over-terror-propaganda-1536744963. Published on the 12/9/2018.