Zhenfeng Gao

The increasing use of video data in training multimodal large language models (MLLMs) raises significant concerns on privacy leakage and copyright violations, highlighting the need for detecting improperly used training videos through membership inference attacks (MIAs). Most existing video MIA m...