![Vasileios Mezaris
Vasileios Mezaris, December 2020
• Whether a video is a carrier of misinformation is not an easy yes/no question; may require
detecting & interpreting manipulations, and understanding intent
• Even relatively simple approaches, such as video fragmentation for reserse Web search, can
have great impact [1]
• AI is the enabler of more elaborate manipulation detection [2], but can also be a risk
[1] D. Teyssou, J.-M. Leung, E. Apostolidis, K. Apostolidis, S. Papadopoulos, M. Zampoglou, O. Papadopoulou, V. Mezaris, "The InVID Plug-in:
Web Video Verification on the Browser", Proc. Int. Workshop on Multimedia Verification (MuVer 2017) at ACM Multimedia 2017, Mountain
View, CA, USA, October 2017.
[2] G. Mercier, F. Markatopoulou, R. Cozien, M. Zampoglou, E. Apostolidis, A. Metsai, S. Papadopoulos, V. Mezaris, I. Patras, I. Kompatsiaris,
"Detecting Manipulations in Video", in book "Video Verification in the Fake News Era", V. Mezaris, L. Nixon, S. Papadopoulos, D. Teyssou
(Eds.), pp. 161-189, Springer, 2019.
[3] "Video verification in the fake news era", V. Mezaris, L. Nixon, S. Papadopoulos, D. Teyssou (Editors), Springer, 2019.
Concluding remarks
• Complete automation is not easy; often, AI+human symbiosis is the key [3]
• New challenges & advances in AI (e.g. Explainable AI) will create new
opportunities for even greater use of AI in verification (of video and more)](https://image.slidesharecdn.com/misinformationontheinternetmezarisdec2020-201217095130/85/Misinformation-on-the-internet-Video-and-AI-9-320.jpg)
Misinformation on the internet: Video and AI
- 1. Misinformation on the internet: Video and AI Vasileios Mezaris CERTH-ITI Presentation at the "Age of misinformation: an interdisciplinary outlook on fake news" webinar 17th December 2020
- 2. Vasileios Mezaris, December 2020 Why focus on video? • Powerful, sensational, persuasive medium • “Believe nothing you hear, and only one half that you see.” (Edgar Allan Poe, Nov. 1845)
- 3. Vasileios Mezaris, December 2020 Can I believe what I see? • Is this video exactly as captured by the camera? (detection) • What has been altered, and how do you know? (explanation) • Why has it been altered? (understanding intent) → Intent is important; are the below edited videos carriers of misinformation? Should they be banned from internet platforms? Video credits: Media Forensics Challenge (MFC) dataset; GRIP, University Federico II, Naples
- 4. Vasileios Mezaris, December 2020 What is the role of AI? • A powerful enabler of detecting and understanding video manipulations • But also a tool for promoting misinformation! (deepfakes; can facilitate Distributed Denial of Truth (DDoT) attacks) Prediction: Tampered Tampering localization / explanation (kind of…)
- 5. Vasileios Mezaris Vasileios Mezaris, December 2020 InVID-WeVerify Verification Plugin: a “Swiss army knife” for video verification A browser plugin to debunk fake news and to verify videos and images >30.000 users • Check prior video use: reverse video search on the Web • Check contextual information: Social- media-based contextual analysis • Keyframe/image inspection by magnifying glass • Check image (keyframe) forensics Free! Get it from: https://www.invid-project.eu/verify Verifying videos: a browser plugin H2020 IA, 2016-2018 Coordinator: CERTH
- 6. Vasileios Mezaris Vasileios Mezaris, December 2020 A CERTH-ITI / InVID webpage for video fragmentation and reverse web search; helps to quicky catch the “easy fakes” (try it!): http://multimedia3.iti.gr/video_ fragmentation/service/start.html Verifying videos: a web service H2020 IA, 2016-2018 Coordinator: CERTH
- 7. Vasileios Mezaris Vasileios Mezaris, December 2020 https://www.youtube.com/watch?v=OVAxQA3gMEo “Video, the crowd in panic flees from Notre Dame in Paris after the attack of an armed man” A video verification example
- 8. Vasileios Mezaris Vasileios Mezaris, December 2020 AI4Media: A European Excellence Centre for Media, Society and Democracy • AI against disinformation is one of our use cases • We are advancing the SoA in AI, in various directions (e.g. explainability, robustness) More AI for verifying videos H2020 RIA, 2020-2024 Coordinator: CERTH
- 9. Vasileios Mezaris Vasileios Mezaris, December 2020 • Whether a video is a carrier of misinformation is not an easy yes/no question; may require detecting & interpreting manipulations, and understanding intent • Even relatively simple approaches, such as video fragmentation for reserse Web search, can have great impact [1] • AI is the enabler of more elaborate manipulation detection [2], but can also be a risk [1] D. Teyssou, J.-M. Leung, E. Apostolidis, K. Apostolidis, S. Papadopoulos, M. Zampoglou, O. Papadopoulou, V. Mezaris, "The InVID Plug-in: Web Video Verification on the Browser", Proc. Int. Workshop on Multimedia Verification (MuVer 2017) at ACM Multimedia 2017, Mountain View, CA, USA, October 2017. [2] G. Mercier, F. Markatopoulou, R. Cozien, M. Zampoglou, E. Apostolidis, A. Metsai, S. Papadopoulos, V. Mezaris, I. Patras, I. Kompatsiaris, "Detecting Manipulations in Video", in book "Video Verification in the Fake News Era", V. Mezaris, L. Nixon, S. Papadopoulos, D. Teyssou (Eds.), pp. 161-189, Springer, 2019. [3] "Video verification in the fake news era", V. Mezaris, L. Nixon, S. Papadopoulos, D. Teyssou (Editors), Springer, 2019. Concluding remarks • Complete automation is not easy; often, AI+human symbiosis is the key [3] • New challenges & advances in AI (e.g. Explainable AI) will create new opportunities for even greater use of AI in verification (of video and more)
- 10. Vasileios Mezaris, December 2020 Questions? Contact: Dr. Vasileios Mezaris Research Director, Head of Intelligent Digital Transformation Laboratory Information Technologies Institute (ITI) / Centre for Research and Technology Hellas (CERTH) Thermi-Thessaloniki, Greece Email: bmezaris@iti.gr, web: http://www.iti.gr/~bmezaris/ This work was supported in part by the EU’s Horizon 2020 programme grant agreements H2020-687786 InVID and H2020-951911 AI4Media