‘Seeing Isn’t Believing’: A Multimodal Dive into Deepfakes

Summary

The rapid advancement of generative AI has revolutionized the creation and manipulation of synthetic media across different use cases. However, this progress has also fuelled the rise of deepfakes; highly realistic audio, image, and video content generated without the consent of the individuals which has potential negative societal impact. These manipulated media artifacts can misrepresent people by making them appear to say or do things they never did, posing significant threats to privacy, trust, and public discourse. Deepfakes have become tools for disinformation, misinformation, online harassment, and even fraud, undermining systems such as facial recognition and video-based authentication (e.g., Know Your Customer aka KYC). As these manipulations grow more convincing, it becomes critical for platforms and systems to detect and mitigate them effectively. This tutorial will provide a comprehensive introduction to the generation, detection, and societal impact of deepfakes from a multimodal perspective. It will equip participants with foundational knowledge of generative techniques, explore cutting-edge deepfake detection methods, and offer hands-on experience with open-source tools. Attendees will also engage in discussion around ethical considerations and real-world deployment challenges, fostering a robust understanding of how to combat synthetic media threats in interactive systems.

Deepfakes have become tools for disinformation, misinformation, online harassment, and even fraud, undermining systems such as facial recognition and video-based authentication (e.g., Know Your Customer aka KYC). As these manipulations grow more convincing, it becomes critical for platforms and systems to detect and mitigate them effectively. This tutorial will provide a comprehensive introduction to the generation, detection, and societal impact of deepfakes from a multimodal perspective. It will equip participants with foundational knowledge of generative techniques, explore cutting-edge deepfake detection methods, and offer hands-on experience with open-source tools. Attendees will also engage in discussion around ethical considerations and real-world deployment challenges, fostering a robust understanding of how to combat synthetic media threats in interactive systems.

Details & Schedule

For details, please contact Abhinav Dhall.
Last updated: 20 May 2025
Website credit: This website style is adopted from the tutorial page.