As generative AI continues to advance, the malicious misuse of AI technologies—particularly deepfakes—has emerged as a serious social issue. Professor Simon Sungil Woo, affiliated with the Department of Computer Science and Engineering, the Graduate School of Artificial Intelligence, and the Department of Applied Data Science at Sungkyunkwan University, is responding to this growing challenge through cutting-edge research on deepfake detection. His work has earned global recognition, ranking 8th worldwide in the deepfake research field according to Google Scholar.

Contrary to the traditional notion that technology remains confined to academic research, the work led by Professor Woo demonstrates how technological innovation can serve as a critical tool for solving real-world problems.

| Could you please introduce yourself?

My name is Simon Sungil Woo, and I am currently affiliated with the Department of Computer Science and Engineering, the Graduate School of Artificial Intelligence, and the Department of Applied Data Science at Sungkyunkwan University. Since joining the university in 2019, I have been actively conducting research in AI security, with a particular focus on deepfake detection.

I immigrated to the United States during my senior year of high school and later returned to Korea. Before returning, I worked as a researcher at NASA’s Jet Propulsion Laboratory (JPL) for approximately nine years. As I continued working, I felt an increasing need to pursue further academic study, which led me to begin my doctoral program later than most. Looking back, it was one of the most worthwhile investments of my life. Today, I feel both busy and fulfilled, sharing new knowledge with students and pursuing the research and work I am truly passionate about.

▲ (From left) Professor Simon Sungil Woo and Sam Altman, Co-founder of OpenAI

| You recently received a Ministerial Commendation from the Ministry of Science and ICT for your research on deepfake detection. Could you tell us more about this work?

I have been conducting research on deepfake detection since 2017. In the past, the misuse of deepfake technology and AI was not widely regarded as a major social issue. However, with the rapid advancement of generative AI, it has now become an increasingly serious problem. The misuse of AI—ranging from information manipulation and fake news to non-consensual exploitation—has grown steadily over time, which I find deeply concerning.

To address this issue, I focus on developing robust and highly generalizable detection technologies capable of identifying various types of deepfakes that continue to evolve. Currently, my research centers on designing deep learning architectures that can efficiently learn from large-scale data. In particular, I leverage large-scale models such as CLIP and DINO to enhance generalization performance. I also apply knowledge distillation techniques, where large, high-capacity models learn first and then transfer that knowledge to smaller models, enabling effective detection even in challenging environments such as low-resolution or highly compressed deepfake content.

The multimodal deepfake dataset (FakeAVCeleb)*, which I first released publicly worldwide, has been widely adopted by researchers around the globe. In addition, in October 2025, our team achieved second place globally at the Deepfake Detection Challenge held during IEEE ICCV, one of the world’s leading computer vision conferences.

* Multimodal Deepfake Dataset (FakeAVCeleb): A core dataset designed to detect more realistic deepfake attacks, containing data in which both audio and video have been simultaneously manipulated.

Based on the research achievements accumulated over a long period, we are confident that our research team is currently at the forefront of deepfake detection technology in Korea. I believe that these achievements, along with our continued efforts to prevent the malicious use of deepfakes and contribute to solving social problems, were highly recognized, leading to this Ministerial Commendation.

| Your research team actively collaborates with organizations such as the Korean National Police Agency. How do these collaborations work in practice, and what outcomes have they produced?

In the past, we have conducted joint research with organizations including the National Forensic Service, Samsung SDS, the Korean National Police Agency, and the Supreme Prosecutors’ Office of Korea. Currently, we are collaborating with the National Police Agency and the National Election Commission.

In particular, we are developing deepfake detection technologies that investigative officers can directly apply in real investigations—technologies that are not only highly accurate but also practical and easy to use. These systems are 100% domestically developed technologies, and we take pride in it as an image and video deepfake detection method created at Sungkyunkwan University.

Through regular meetings with the Korean National Police Agency, we continuously identify challenges encountered in real investigations and refine our models to provide more practical support. Beyond Korea, we are also collaborating with the Düsseldorf Police Department in Germany, where our detection models have been deployed as APIs and are currently being tested and utilized.

In preparation for the upcoming local elections in June this year, we have also begun cooperation with the National Election Commission to proactively prevent the misuse of deepfake technologies. Seeing our research contribute to building a safer and more trustworthy society is deeply meaningful to me.

| Given the rapid pace of technological change in deepfake-related research, you must have faced many challenges. What difficulties have you encountered, and how have you addressed them?

AI technology is evolving at a remarkably rapid pace, and generative AI, in particular, is increasingly being misused in deepfake creation. When new deepfake techniques emerge that were not included in a model’s training data, detection performance can drop significantly. This makes the continuous collection of new data, retraining of models, and development of new architectures are one of the key challenges in this field.

To overcome these obstacles, we apply techniques such as continual learning, domain adaptation, various data augmentation methods, and advanced face recognition techniques to improve detection performance. More recently, we have also leveraged the tendency of generative AI–produced images and audio to exhibit repetitive patterns in specific frequency domains, continuously updating our models to improve detection accuracy.

Our most pressing challenge is developing detection techniques that maintain high performance in unpredictable real-world environments. In our lab, we continue to advance research aimed at detecting deepfakes that are actively misused in real-life scenarios.

| Your participation in the special exhibition “Immortal Heroes of the Korean War, Returning as Youth” was particularly notable. How did you become involved, and what did this project mean to you as a researcher?

▲ Restored image of General Kim Duman using AI face restoration (GFP-GAN), Source: Ministry of Patriots and Veterans Affairs

We were approached by the Ministry of Patriots and Veterans Affairs on the occasion of the 70th anniversary of the Korean War armistice, with a request to explore whether AI technology could be used to restore damaged and faded photographs of Korean War heroes. Recognizing the profound significance of this work, students and I were honored to participate.

As a researcher, it was deeply rewarding to have the chance to use the technologies we have developed to pay even a small tribute to the dedication and the sacrifices of heroes who devoted their lives to the nation. Moreover, seeing that the restored photographs brought great joy to the bereaved families made the project all the more fulfilling.

| Your research consistently emphasizes solving social problems through technology. What values guide you most when choosing research topics?

When determining research topics, I prioritize the following criteria:

  1. Does technology have the potential to benefit society beyond its economic value?

  2. Is it meaningful, and can it genuinely help people?

  3. Is it a topic that my research team and I can enjoy working on?

  4. Is it a relatively unexplored or emerging research area?

  5. Is it a technology that will be essential in the future?

While academic value is important, I believe that technology development should ultimately serve people and society as a whole. With the rapid advancement of AI, we are already witnessing how human and social values may be altered or even undermined. As the risks and negative impacts of AI continue to grow, I find it increasingly concerning, which has led me to focus on AI technologies that can contribute positively to society while upholding human-centered values.

| You previously worked as a researcher at NASA’s Jet Propulsion Laboratory. Are there any memorable experiences from that time that have influenced your current work?

My first space mission project was the Mars Reconnaissance Orbiter (MRO) in 2005. I worked on developing the CFDP file transfer protocol and Delay-Tolerant Networking (DTN) to transmit image data captured on Mars back to Earth. Knowing that a spacecraft I helped develop traveled to Mars—and continues to operate even more than 20 years beyond its original design life—is truly astonishing.

▲ Graphic image of the Mars Reconnaissance Orbiter, Source: NASA

Beyond any material value, simply having the opportunity to contribute to a project dedicated to the scientific advancement of humanity was an incredibly valuable experience. It was also a great honor to collaborate with outstanding engineers, scientists, and technicians at NASA. The collaboration practices, approaches to teamwork and harmony, methods of tackling research problems and the ways of writing research papers and proposals that I learned there have been immensely beneficial to me across many aspects of my work today.

| Are there any new research areas you hope to explore in the future?

I am interested in further exploring AI security, human-centered AI technologies, and new AI foundation models that respect and uphold human values.

▲ Professor Simon Sungil Woo restoring photographs with students, Source: The Electronic Times

| Finally, do you have a message for students interested in joining your DASH Lab?

Students who wish to move beyond simply pursuing popular, profit-driven AI technologies—such as improving the performance of large language models—and instead want to research and develop truly meaningful AI technologies that address fundamental limitations are welcome to reach out. Above all, I hope students who are sincere, responsible, and passionate about research will apply, as these qualities are essential to building a strong and positive research lab.

>Dash Lab Website

Interview: Kim Eunseo