FSU faculty members Dr. Shuyuan Metcalfe (School of Information) and Dr. Xiuwen Liu (Department of Computer Science) are currently working on exciting research to identify “deepfake” images. The recently awarded Project Aletheia from the U.S. Air Force (USAF) allows FSU investigators to develop machine learning (ML) models that will detect digitally manipulated images. The research team is also working to develop a secure cyberinfrastructure to facilitate the user experience on this artificial intelligence (AI) platform.
The objective of Project Aletheia is to study and detect adversarial deepfake image manipulation. With technology advancement that goes far beyond Photoshop, Internet users can now easily modify photos to confuse perception and public opinion. Dr. Metcalfe, whose research focuses on trusted human-computer interaction as well as computer-mediated deception and disinformation, explained “Deepfake images, voice, videos, or any types of digital artifacts can be modified as a type of computer-mediated deception that aims to confuse individuals and society.”
When asked about the impact of this kind of deception, she stated, “The power of disinformation and deepfake manipulation cannot be ignored. Sophisticated generative adversarial networks (GANs) are being utilized by ill-intentioned individuals. Misinformation twists the truth and also has the potential to polarize communities.”
The team’s research has progressed significantly since it began; Dr. Metcalfe shared that the recent midterm program review with USAF was quite successful as the team was able to demonstrate the effectiveness of several machine learning models. Next, the team will begin developing a “truth disclosure” platform that allows users to check image authenticity.
Dr. Metcalfe discussed how her team is doing this, stating, “We are testing and evaluating different models, and designing an approach to utilize strategic machine learning that uses artificial intelligence to identify digitally enhanced images.” The systems and infrastructure for ML models’ training, testing, and experimentation has been built through the utilization of high-performance computing, supported by the FSU Research Computing Center. “What is more intriguing,” Dr. Metcalfe shared, “is that the project team is developing a cyberinfrastructure that caters to the user experience—allowing Aletheia users to experience artificial intelligence and high-performance computing in near real-time.”
Dr. Metcalfe also expressed her gratitude for the support she has received throughout the project, saying, “We appreciate the support we have received from Paul van der Mark and the staff at the Research Computing Center, as well as Shawn Banner, Kyle Muller, and the technical support team at the College of Communication and Information (CCI). They have been crucial in helping to set up a cyberinfrastructure that’s suitable and secure under the University policy so we can continue our research and development work.”