The Media Verification team (MeVer) emerged as a result of our research in a number of projects in the areas of social media mining, search and online disinformation.We made our first steps in this area already in 2013 in the context of the SocialSensor project, where our collaboration with experts from the field of Media and Journalism, soon led us to acknowledge the rise of misleading media content in social media platforms such as Twitter. In the next years, we conducted research in the context of two follow-up projects, REVEAL and InVID, and developed technical components that assisted with the mitigation of different aspects of the media-based disinformation challenge, including among others a toolbox of image forensics approaches and a web-based service that assists with image verification, a tweet verification assistant, a service for analyzing the online context of a video, and a service for the retrieval of near-duplicate videos.
In addition to the aforementioned tools, we have been very actively dealing with the development of new benchmarking resources and methods. In 2015 and 2016, we organized the Verifying Multimedia Use challenge in the context of the MediaEval Benchmarking Initiative, which was accompanied by the release of an annotated corpus of tweets (categorized as "fake" or "real"). Next, to support the real-world benchmarking needs of the image tampering localization problem, we developed the Wild Web tampered image dataset, which has been a valuable resource for image forensics researchers worldwide. Our recent dataset releases include the Fake Video Corpus and the FIVR-200K datasets, the first aiming to help researchers better understand the nuances of the video verification challenge, the latter focusing on the problem of near-duplicate video retrieval in relation to breaking news incidents.
While we continue to support and refine our technologies in this rapidly evolving area, we are currently also trying to better understand and tackle the challenge of AI-generated fake multimedia (aka deepfake) in the context of the WeVerify project. This is an increasingly complex challenge given the proliferation and constant improvement of deep-learning based methods for generating realistic media content of high fidelity and the continuous evolution of means and strategies of online misinformation. We'll be posting our findings and hopefully useful resources on this website.
Key team members include Senior Researcher Nikos Sarris (media analysis and verification tools), researcher Olga Papadopoulou (context aggregation and analysis, web service engineering), PhD candidate George Kordopatis-Zilos (near-duplicate retrieval, deep learning, language models), post-doc researcher Chryssanthi Iakovidou (image forensics), researcher Polychronis Charitidis (deepfakes), and web master Lazaros Apostolidis (front-end development). The team is led by Senior Researcher Symeon (Akis) Papadopoulos and is part of the Multimedia Knowledge and Social Media Analytics (MKLab), directed by Senior Researcher Yiannis Kompatsiaris.
We would also like to acknowledge the valuable contributions of our past colleagues, Markos Zampoglou (image and video forensics, deep learning), Katerina Andreadou (media crawling and search), Christina Boididou (tweet verification assistant), and George Rizos (graph-based account classification).
In our research projects, we've worked with numerous researchers and experts in the areas of Media, Journalism, Web Mining and Multimedia. These collaborations were an inspiration for us and helped broaden our perspective and deepen our understanding of the challenges we are facing. As a minimum token of our appreciation, we would like to acknowledge them here (in alphabetic order): Giulia Boato (University of Trento), Petter Bae Brandtzaeg (SINTEF), Kalina Bontcheva (The University of Sheffield), Duc-Tien Dang-Nguyen (University of Bergen), Zhiwei Jin (Chinese Academy of Sciences), Vasilis Mezaris and Lampis Apostolidis (Centre for Research and Technology Hellas), Stuart E. Middleton (IT Innovation Centre), Nic Newman (Reuters Institute for the Study of Journalism), Ioannis Patras (Queen Mary University, London), Danae Tsampouraki and Stratos Tzoannos (Athens Technology Center), Steve Schifferes (City, University of London), Jochen Spangenberg and Ruben Bouwmeester (Deutsche Welle), Denis Teyssou (Agence France Presse), Andreas Veglis, Charalampos Dimoulas, Rigas Kotsakis and Dimitris Giomelakis (Aristotle University of Thessaloniki).
Work with Us
We are always open in new collaborations and partnerships. If you are a researcher and have an interesting idea that we could pursue together, please get in touch. If you are a graduate student or researcher and are looking for opportunities to carry out research in these areas, please get in touch.
If you are a company looking for solutions in the areas of online media analysis and verification, we are open to licensing our technologies and services, or discussing possibilities of contracted research tailored to your needs.
Given that funding is necessary for carrying out our activities, we are always interested in participating in collaborative project proposals (e.g. for EC-funded or national research and innovation projects).