About DeFake Project

The DeFake Project is a research initiative focused on developing effective, usable, and interpretable deepfake detection tools and advancing the field of digital media forensics. Our mission is to create innovative technologies that empower journalists, intelligence analysts, and law enforcement to identify and verify the authenticity of digital content.

Our Origins

The DeFake Project is a research lab located at the ESL Global Cybersecurity Institute at Rochester Institute of Technology. Our work is conducted in collaboration with partners from the School of Journalism and New Media at University of Mississippi and Michigan State University. We are honored to be one of the eight winners of the Knight Foundation’s AI and the News Open Challenge, recognizing our innovative approach to addressing the challenges posed by deepfakes in journalism and beyond.

The Challenge We Address

The rapid advancement of deepfake technology has significantly lowered the barriers to creating convincing, manipulated video content. This poses several critical challenges, including the potential for disinformation campaigns targeting democratic processes, risks to the reputations of individuals and organizations, and erosion of public trust in visual media and news sources.

Our Approach

The DeFake Project takes a usability-centered approach, emphasizing the development of tools that are not only effective but also intuitive and interpretable for end-users. We combine user studies with professionals in the field and the development of novel detection methods. Our work focuses on creating user-friendly tools tailored to the needs of journalists, analysts, and law enforcement, while also developing and deploying our own advanced detection algorithms that provide clear, understandable results. We conduct ongoing research to improve the overall field of digital media forensics, with a particular emphasis on making complex AI-driven detection methods more transparent and interpretable. Our methods integrate advanced computer vision analysis and metadata examination to provide comprehensive and practical solutions for deepfake detection and content verification. Additionally, we work closely with partners to provide expert deepfake analysis, ensuring that our tools and methodologies meet the real-world needs of professionals dealing with potentially manipulated content.

Community Engagement

We frequently engage with journalism and law enforcement communities through various initiatives. This includes participating in and organizing workgroups focused on deepfake detection challenges, providing training sessions on deepfake and deepfake detection technologies, and delivering talks and presentations at industry events and conferences. These activities allow us to share our expertise, gather valuable insights from practitioners, and ensure our research remains aligned with real-world needs while continuously improving the usability and interpretability of our tools.

Project Features

The DeFake Project offers efficient deepfake detection technology that enables quick and accurate identification of manipulated video content. Our user-centric design approach ensures that our tools meet the specific needs of journalists, intelligence analysts, and law enforcement professionals, with a strong focus on providing clear, interpretable results. We also provide educational resources to help users understand the complexities of deepfakes and digital manipulation, emphasizing the importance of human judgment in conjunction with our AI-powered tools. As part of our commitment to advancing the field, we regularly contribute to the scientific community through peer-reviewed publications and conference presentations, often highlighting our findings on improving the usability and interpretability of deepfake detection methods.

Acknowledgements

The DeFake Project is made possible through the support of Rochester Institute of Technology, our partner institutions, the Knight Foundation, the National Science Foundation (NSF), and various other research grants. We extend our gratitude to our academic partners, industry collaborators, and the open-source community for their contributions to this vital work in making deepfake detection more effective, usable, and interpretable.