大阪大学 / 工学部電子情報工学科
Digital Watermarks in Text-to-Image Generation Models
Conducted in-depth research on the validity and effectiveness of membership inference attacks in text-to-image generation models. Published a paper highlighting the need for more robust watermarking technologies in text-to-image generation models due to the ineffectiveness of current methods. Presented research findings at the prestigious CSEC conference, engaging with industry experts and peers to discuss the implications of membership inference attacks and watermarking technologies in text-to-image generation models. Contributed to the ongoing discourse on model privacy and security, emphasizing the importance of developing advanced watermarking techniques for text-to-image generation systems.