Membership Inference Attacks Against Video Large Language Models

Abstract

Video large language models (VideoLLMs) are increasingly trained or instruction-tuned on large-scale video--text corpora collected from heterogeneous sources, raising an immediate privacy question: can an external auditor determine whether a particular video was used during training? While membership inference attacks (MIAs) have been studied extensively for classifiers and, more recently, for text and image generation models, the VideoLLM setting remains unexplored. This setting is challenging

Categories

Framework mappings

Suggested from the entry's categories.

@misc{song2026membership,
  title = {{Membership Inference Attacks Against Video Large Language Models}},
  author = {Wei Song and Yuxin Cao and Ziqi Ding and Yi Liu and Gelei Deng and Yuekang Li},
  year = {2026},
  month = apr,
  eprint = {2604.27002},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2604.27002}
}