Researchers at the University of Hull have developed a new method for detecting deepfakes, based on analyzing the reflection of light in the eyes using astronomical instruments. The method has shown promising results, but requires further improvement.
Detecting deepfakes
An innovative approach to identifying fakes
Popular wisdom says that the eyes are the mirror of the soul. However, recent research has shown that they can also be an effective tool for identifying deepfakes. Scientists propose using techniques used in astronomy to study galaxies to analyze the reflection of light in the eyes in the images. This approach was developed by Adejumoke Owolabi, a master's student at the University of Hull in the UK.
Science and Technology Collaborations
Owolabi worked in tandem with Kevin Pimblett, professor of astrophysics and director of the Center for Advanced Data Science science, AI and modeling. Their team conducted a comparative analysis of real photographs of people and images generated by artificial intelligence. Two astronomical methods were used to study light reflections in the eyes of both image categories: the Gini coefficient and the CAS system.
Astronomical Methods in the Service of Image Recognition
The Gini coefficient is used to estimate the concentration of light in galaxy images based on pixel analysis. Coefficient values from 0 to 1 allow you to determine the structure of the galaxy - smooth or crowded, which is typical for elliptical or spiral galaxies, respectively. The CAS system, in turn, helps astronomers measure the distribution of light in galaxies to determine their morphology.
Research results
The research team used both tools to compare left and right eyeballs in real photographs and AI-generated images. Although the CAS system did not show sufficient effectiveness in detecting deepfakes, the Gini coefficient showed significant differences.
Research has shown that if the reflections of light in both eyes match, then the image is highly likely to be genuine. Inconsistent highlights, on the other hand, may indicate a deepfake.
Prospects and limitations of the method
Professor Pimblett emphasizes that this method is not an absolutely reliable means of identifying fake images. There is a possibility of both false positive and false negative results. However, this approach provides a framework and strategy for further development of deepfake detection methods in an ever-changing technological environment.
Glossary
- Deepfake is a technology for synthesizing an image or video of a person using artificial intelligence
- The University of Hull is a higher education institution in the UK founded in 1927
- The Gini coefficient is a statistical indicator used to measure the skewness of a distribution
- CAS system - a method for analyzing galaxy images based on measuring concentration, asymmetry and clumping
- AI (Artificial Intelligence) - technology for creating intelligent machines and computer programs
Links
Answers to questions
How can eyes help detect deepfakes?
What astronomical instruments are used for image analysis?
How accurate is deepfake detection using eye analysis?
Who did the research on deepfake detection?
Why is the Gini coefficient more effective than the CAS system?
Hashtags
Save a link to this article
Discussion of the topic – A revolution in deepfake detection: from galaxies to eyes
The article describes a new method for detecting deepfakes by analyzing the reflection of light in the eyes, using tools previously used to study galaxies.
Latest comments
8 comments
Write a comment
Your email address will not be published. Required fields are checked *
Alicia
Wow, this is so interesting! 🤓 I would never have thought that astronomical instruments could be used to identify deepfakes. It's amazing how science can be applied in the most unexpected areas!
Hans
Yes, Alicia, that's really impressive! But I worry that criminals can quickly adapt and learn to circumvent this method. How long do you think it will be effective? 🤔
Sophie
Hans, you're right, it's an arms race. But I think any new detection method is a step forward. Even if it's not perfect, it forces deepfake creators to work harder, which can slow their spread. 💪
Giovanni
I agree with Sophie. By the way, I was amazed how accurately the Gini coefficient can determine the difference in glare. I wonder if this technology can be applied to video? 🎥
Viktor
Ha, another useless technology. Instead of wasting time fighting windmills, it would be better to do something really important. Still, scammers will always find a way to bypass the protection.
Alicia
Victor, I don't agree. Any progress in deepfake detection is important. This can help protect people from scams and misinformation. Don't give up just because the task is difficult! 💡
Hans
I support Alicia! Giovanni, your idea for the video is interesting. Maybe we should write to the researchers and suggest developing the technology in this direction? This could be a breakthrough in the fight against deepfakes on social networks! 📱
Sophie
Great idea, Hans! I would love to participate in such a project. Imagine how this could change the landscape of fake news and disinformation on the Internet. Does anyone know how to contact these scientists? 🕵️♀️