Suggestions
Hany Farid
Professor of Computer Science specializing in image analysis and digital forensics
Hany Farid is a distinguished professor at the University of California, Berkeley, with a joint appointment in Electrical Engineering & Computer Sciences and the School of Information.1 His expertise lies in digital forensics, forensic science, misinformation, image analysis, and human perception.12
Academic Background
Farid received his undergraduate degree in Computer Science and Applied Mathematics from the University of Rochester in 1989, followed by an M.S. in Computer Science from SUNY Albany, and a Ph.D. in Computer Science from the University of Pennsylvania in 1997.2 After completing a two-year post-doctoral fellowship in Brain and Cognitive Sciences at MIT, he joined the faculty at Dartmouth College in 1999, where he remained until 2019.23
Professional Achievements
Farid has received several prestigious accolades, including:
- Alfred P. Sloan Fellowship
- John Simon Guggenheim Fellowship
- Fellow of the National Academy of Inventors2
He is widely recognized as a pioneer in the field of digital forensics and has been called the "father" of digital image forensics.4
Current Roles and Affiliations
At UC Berkeley, Farid holds multiple positions:
- Professor in Electrical Engineering & Computer Sciences and the School of Information
- Member of the Berkeley Artificial Intelligence Lab
- Member of the Center for Innovation in Vision and Optics
- Member of the Vision Science program23
Additionally, he serves as:
- Senior Advisor for the Counter Extremism Project
- Director on the Board of Directors for the Cyber Civil Rights Initiative
- Advisory Board Member for the Human Rights Center at UC Berkeley School of Law
- Content Advisory Council member for TikTok3
Research and Expertise
Farid's research focuses on developing mathematical and computational techniques to detect various forms of tampering in photos, videos, audios, and documents.5 He is particularly known for his work on:
- Detecting digitally manipulated images, including deepfakes
- Analyzing the human perception of manipulated images
- Addressing issues related to misinformation and disinformation14
His expertise has led him to consult for intelligence agencies, news organizations, courts, and scientific journals seeking to authenticate the validity of images.4
Hany Farid's work continues to be crucial in the evolving landscape of digital media and its impact on society, particularly in the areas of misinformation and digital forensics.
Highlights
I am fielding a lot of questions about the attached montage of these "Swifties for Trump" photos (which I've converted to grayscale to reduce the spread of this misleading montage).
The six photos in the upper left are too small to forensically analyze but they have the telltale visual signs of being AI-generated.
The photo in the upper right appears to be authentic.
Based on the results of two models trained to detect traces of AI generation, the image in the lower right (left panel) is AI generated, while the image in the lower right (right panel) appears to be authentic.
For the two seemingly authentic photos, our models find no traces of AI, and a shadow and perspective analysis reveals no inconsistencies.
Combining real and fake photos into a single montage is a clever/devious way of pushing misleading information.
I am fielding a lot of question about this side-by-side comparison of images that is making the rounds on X: https://lnkd.in/dbCA9kW7.
I analyzed this image with our GetReal Labs model trained to distinguish real from AI-generated images and found no evidence of AI-generation or manipulation.
Shown below, however, is a visualization of the image/sensor noise in the historical photo. In an ideal imaging device, the pixel values of the digital image would accurately reflect the amount of light recorded by each photo detector. Real devices have imperfections, however, and these imperfections introduce noise in the image. The resulting noise pattern is typically random, fluctuating from image to image but generally uniform across the image.
Below you can see a much larger noise pattern (brighter regions) for three main parts of the image (the two people and the flag). This would not happen naturally and is a tell-tale sign of photo manipulation. While I can’t say exactly how this image was created, it is most likely that this is an old-school Photoshop manipulation in which several photos were composited together.
#deepfakes #cheapfakes
