Enterprise Organization of National Program of Excellence in Software

Along with the advancement of generative AI, the issue of deepfakes exploited through AI is becoming increasingly serious. Professor Simon Sungil Woo, who is affiliated with the Department of Computer Science and Engineering, the Graduate School of Artificial Intelligence, and the Data Science Convergence Major at our university, is responding to this social trend through deepfake detection research. He is revealing his prominence in the field, ranking 8th globally in the deepfake category according to Google Scholar.

Contrary to the past notion that technology is limited to academic research, the technology demonstrated by Professor Simon Sungil Woo serves as the core key to solving practical problems. In this 579th People Focus, we listen in detail to his philosophy as an artificial intelligence researcher and educator, and raise the question of what direction technology should be researched in.

| Hello, Professor. Please introduce yourself briefly.

I am Simon Sungil Woo, currently affiliated with the Department of Computer Science and Engineering, the Graduate School of Artificial Intelligence, and the Data Science Convergence Major at Sungkyunkwan University. Since arriving at Sungkyunkwan University in 2019, I have been diligently conducting research related to artificial intelligence security, particularly deepfake detection. I immigrated to the United States during my senior year of high school and lived there before returning to Korea. Before returning to Korea, I worked as a researcher at the NASA Jet Propulsion Laboratory for about nine years. As I continued my career, I felt the need to study more and started my doctoral program late, but looking back now, I think it was the greatest investment of my life with no regrets. Currently, I am spending busy yet happy days being able to share new knowledge with students and conduct the research and work I want to do.

▲ (From left) Professor Simon Sungil Woo, OpenAI founder Sam Altman

| You recently received a Minister's Citation from the Ministry of Science and ICT for your research on deepfake detection technology. Could you explain that research in detail?

I have been conducting research related to deepfake detection since 2017. In the past, instances of deepfakes and the misuse of artificial intelligence were not considered major social problems, but with the development of generative AI technology, they are becoming increasingly significant issues. It is regrettable that cases of AI misuse in many fields, such as information manipulation, fake news, and non-consensual sexual imagery of acquaintances, are increasing over time.

To solve this, I am researching detection technology with high robustness and generalization performance that can detect various types of evolving deepfakes with high performance. Currently, I am researching deep learning architectures that can detect deepfakes by efficiently learning from large amounts of data. In particular, I am using large deep learning models such as CLIP and DINO to increase the generalization performance of detection. Additionally, I am conducting research to detect deepfakes in various environments, especially low-resolution or compressed ones, by utilizing Knowledge Distillation techniques, where a large and smart model first learns from low-quality models and then transfers that content to a smaller model. Results have been achieved, such as many researchers around the world utilizing the multimodal deepfake dataset (FakeAVCeleb)* that I released for the first time in the world. Also, our team achieved the feat of ranking 2nd in the world at the deepfake detection challenge held at IEEE ICCV, a world-renowned computer vision academic conference, in October 2025.
*Multimodal deepfake dataset (FakeAVCeleb): Core data for detecting more realistic deepfake attacks, including data where voice and video are manipulated simultaneously.

Based on the research achievements accumulated over a long period, I am confident that our research team is the best in the field of domestic deepfake detection technology. I believe I received this Minister's Award because these research achievements, along with my efforts to prevent the misuse of deepfakes and contribute to solving social problems, were highly evaluated.

| I understand that your research team does not stop at research but is contributing to solving deepfake issues by cooperating with various organizations such as the Korean National Police Agency. How is this collaboration actually taking place, and could you introduce any achievements?

In the past, we conducted joint research with the National Forensic Service, Samsung SDS, the Korean National Police Agency, and the Supreme Prosecutors' Office. Currently, we are cooperating with the Korean National Police Agency and the National Election Commission. In particular, we are developing technology for the Korean National Police Agency that is easy to use with high accuracy and utility so that investigators can apply it to actual investigations. I am proud to say that this technology is 100% domestic technology and is an image/video deepfake detection method developed at Sungkyunkwan University. Furthermore, through periodic meetings with the Korean National Police Agency, we are continuously improving the model by identifying specific difficulties in investigations and how we can provide more practical help. In addition to the domestic Korean National Police Agency, we are also collaborating with the Dusseldorf Police Department in Germany, and we have turned the detection model we created into an API, which the German police are currently testing and utilizing.

Also, in preparation for the local elections in June 2026, we have begun cooperating with the National Election Commission to prevent the misuse of deepfakes in advance. It is personally meaningful that the detection technology we created is contributing to making a better society.

| I imagine there were many difficulties while researching a field where technology develops as fast as deepfakes. I am curious about the obstacles you faced during the research process and how you have overcome them.

AI technology is evolving very rapidly. In particular, generative AI technology is being applied to deepfakes and is being widely misused. In other words, when a new deepfake technique that was not used in the training data emerges, a problem arises where the performance of existing models drops significantly. Therefore, the process of continuously collecting new data, training models, and creating new models is an obstacle. To solve this, we are improving detection performance by applying Continual Learning, Domain Adaptation Methods utilized in machine learning, and various Data Augmentation techniques and high-performance face recognition techniques. Also, we are updating the model to continuously increase detection performance by utilizing the tendency of images or voices created by recent generative AI to have repetitive patterns in specific frequency domains.

The biggest task at hand is to create a detection method that shows high performance even for deepfakes in the real world, which are difficult to predict. I and our lab are continuing research on detecting deepfakes misused in reality by utilizing the aforementioned technologies.

| Among your career highlights, your participation in the Special Photo Exhibition: Immortal Heroes of the June 25 War, Returning as Young Men was impressive. What was the motivation for participating in that exhibition and what was the significance of this project to you as a researcher?

▲ A photo of General Kim Doo-man restored using AI face restoration technology (GFP-GAN) and Face Restoration. Source = Ministry of Patriots and Veterans Affairs

By chance, the Ministry of Patriots and Veterans Affairs inquired whether it would be possible to use artificial intelligence technology to restore damaged and faded photos of June 25 heroes in celebration of the 70th anniversary of the end of the June 25 War. It was such a meaningful task that the students and I were happy to participate. As a researcher, it was a project I was very happy to carry out as it became an opportunity for the technologies we developed to respond, even if only a little, to the hard work and sacrifice of the heroes who nobly gave their lives for the country. Also, it seems the reward was even greater because the restored photos were able to bring great joy to the bereaved families, as shown below.

-Words from a bereaved family member of a June 25 hero in a photo restored at the time-"

I am immensely grateful to the students of the Department of Computer Science and Engineering and the Graduate School of Artificial Intelligence. In particular, Professor Simon Sungil Woo, you worked so hard. I tried to find a way to convey my gratitude but couldn't, so I am leaving words of thanks here to express my appreciation. (Omitted) Last March, the Ministry of Patriots and Veterans Affairs and the students of the Department of Computer Science and Engineering and the Graduate School of Artificial Intelligence worked with blood, sweat, and tears to restore the only remaining photo in the world so naturally, prettily, and beautifully. Only one photo remained from my father's days as a student in the 8th class of the Korea Military Academy 70 years ago, which had faded, and it was barely sent home after requesting it from the academy's civil service office. That one photo of my father who died in battle is my number one precious treasure that cannot be exchanged for money or anything else. I am beyond grateful that you restored such a photo into a natural color photo so vividly. I would like to take this opportunity to express my sincere gratitude and thanks to Sungkyunkwan University and the students who worked so hard."

>Go to see the article related to the project

| Looking at your research, there is a common significance of solving social problems through technology. What are the criteria or values you prioritize most when deciding on a research topic?

The criteria I use to decide on a research topic are as follows.

1. Can the technology be helpful to society beyond economic value?

2. Is it a technology that is meaningful and can truly help people?
3. Is it a topic that researchers, including myself, can enjoy researching?
4. Is it a new field that has not been researched much before?
5. Is it a technology that will be essential in the future?

Academic value is important, but at the same time, I believe it is necessary to develop technology that fundamentally helps people and further helps society as a whole. In particular, we can see even now that human and social values can be fluctuated and damaged by the development of artificial intelligence. Also, I am very worried as the harms and risks of artificial intelligence are increasing. Therefore, as a researcher and engineer, I have a great interest in and prepare for research and development of artificial intelligence that can contribute to society while encompassing human values.

| You also have experience working as a researcher at the NASA Jet Propulsion Laboratory. Please introduce any memorable episodes or experiences from that process that influenced your current research.

The Mars Reconnaissance Orbiter (MRO)* research I participated in in 2005 was my first space mission project. I researched and developed the CFDP file transfer protocol and DTN (Delay-Tolerant Networking) that transmit photo files taken on Mars to Earth. It seems truly amazing that the satellite I participated in developing went to Mars and that it is still operating today, 20 years later, far exceeding its design life.

* Mars Reconnaissance Orbiter (MRO): A spacecraft designed to search for the presence of water on Mars and support Mars exploration missions, providing evidence such as safe landing site selection for Mars landers and additional evidence of water flowing on the Martian surface.

▲ Graphic image of the Mars Reconnaissance Orbiter. Source = NASA

Regardless of material value, the fact that I could participate in a project for the development of mankind scientifically was truly a precious experience. It was also a great honor to be able to research with truly outstanding engineers, scientists, and technicians at NASA. Working with them, I learned how to work when multiple people carry out a task together. Currently, I am conducting research that is completely different from what I did at NASA, but the collaboration methods, ways of harmony, approach to research problems, and how to write research papers and proposals that I learned there are providing great help to me in many ways today.

| Are there any research fields you would like to newly challenge in the future, or research topics you are currently preparing with interest?

I would like to research artificial intelligence security and, as mentioned before, artificial intelligence technology that treasures and respects human values, as well as new artificial intelligence foundation models.

| In addition to research activities, you are performing various roles as an educator and lab head. I am curious about what you think a 'good lab' looks like.

I believe that in a good lab, all participating researchers must have a sense of responsibility for their assigned tasks, both humanly and professionally. This is a truly important part everywhere as it is a consideration and promise to each other. In any organization, an individual cannot do anything alone, no matter how outstanding they are. It seems that better results come out when individuals gather and create synergy. Therefore, members should respect and be considerate of each other and not be selfish.

I hope students think of study and research as an investment in their own future. Since the process is not easy, according to what I have seen, it is not the smart students who succeed, but the students with persistence who work hard that succeed in lab life and society. Therefore, if there are many responsible and persistent students, I believe performance will naturally follow and it will become a good lab on its own.

▲ Professor Simon Sungil Woo restoring photos with students. Source = Electronic Times

| Lastly, please say a word to students who are interested in your lab, the DASH Lab.

I hope students who want to go a step further beyond artificial intelligence technology that is simply of high interest now and can make a lot of money (e.g., how to increase the performance of LLM?) and research and develop truly meaningful artificial intelligence technology (e.g., how to fundamentally solve the problems of LLM?) will contact me. Also, as mentioned before, I hope students who are kind, responsible, and will research hard, which is necessary to create a good lab, will apply.

>Go to read the DASH Lab visit article of Professor Simon Sungil Woo

>DASH Lab Homepage