NCKU Proposes Deepfake Detection for 2024 Taiwan Lantern Festival, Enhancing Multimedia Security with Green Computing
Deepfake, utilizing AI technology to forge or swap faces for criminal activities or attempts to influence public opinion, has become increasingly prevalent, impacting international socio-political affairs and individuals' daily lives, challenging the traditional belief in "seeing is believing," with significant ramifications. The NCKU team, with expertise in data science, statistics, low-power computing, and brainwave experiments, pioneers a novel and effective deep learning framework in the field of multimedia image security. This comprehensive AI-driven image defense solution not only applies to public life but also receives international recognition and expectations in the fields of Earth and space sciences' remote sensing imagery.
Associate Professor Chih-Chung Hsu, the project's lead, notes that current verification units do not directly employ detection technology, ultimately relying on human judgment for multimedia authenticity. The team's current endeavor is to "let people help AI, and then let AI help people." He emphasizes that with advancing technology, deepfake visuals are becoming increasingly realistic and challenging for human eyes to discern, while evasion of AI detection attacks continues to evolve, increasing the difficulty of deepfake video recognition. The team's project website for deepfake detection is continuously being optimized in terms of both hardware and software. In the future, they hope to further upgrade detection and defense capabilities by understanding the human brain's mechanism for interpreting false information, as well as accelerate low-power development to overcome current computational limitations and deploy the technology to more devices.
After a 16-year hiatus, the Taiwan Lantern Festival returns to Tainan, and three interdisciplinary teams from NCKU present their outstanding research achievements at "Illuminate Shalun." The third showcase on March 6th is led by Associate Professor Chih-Chung Hsu from the Institute of Data Science and Statistics, focusing on multimedia information security. The project, titled "Green Computing and Sustainable Learning-Based False Multimedia Message Recognition," not only explores new dimensions in deepfake detection but also creates new research fields and possibilities for industry innovation. In terms of maintaining the integrity and security of cyberspace, the team prioritizes promoting a peaceful and just social environment. Their sustainable development goals include quality education (SDG4), affordable and clean energy (SDG7), industry innovation and infrastructure (SDG9), and peace, justice, and strong institutions (SDG16).
The NCKU interdisciplinary project exhibition is located in the "Illuminate Shalun" exhibition area of the 2024 Taiwan Lantern Festival, adjacent to Guiren 18th Road, open from 1:00 PM to 9:00 PM daily. For more transportation information, please visit the "2024 Taiwan Lantern Festival in Tainan" official website or refer to the traffic information of the Greater Tainan Exhibition Center and "Good Traffic Operation in Tainan" Facebook page.
As the 2024 Taiwan Lantern Festival comes to a close, Associate Professor Chih-Chung Hsu from the Institute of Data Science and Statistics at NCKU leads a project team in unveiling the globally pioneering "Deepfake Detection" technology.
Bringing together expertise in data science and statistics along with low-power computing technology and brainwave experiments, the NCKU team has pioneered a novel and effective application of deep learning architecture in the field of multimedia image security. This architecture is capable of simultaneously defending against deepfake manipulation and countering attacks.
After a hiatus of 16 years, the Taiwan Lantern Festival returns to Tainan, where three interdisciplinary teams from NCKU showcased their remarkable research achievements in the "Lighting Up Shalun" exhibition. The third exhibition will run until March 10th, welcoming visitors to seize the opportunity to visit before the weekend.