Private Attribute Inference from Images with Vision-Language Models

arXiv:2404.10618v1 Announce Type: cross
Abstract: As large language models (LLMs) become ubiquitous in our daily tasks and digital interactions, associated privacy risks are increasingly in focus. While LLM privacy research has primarily focused on the leakage of model training data, it has recently been shown that the increase in models' capabilities has enabled LLMs to make accurate privacy-infringing inferences from previously unseen texts. With the rise of multimodal vision-language models (VLMs), capable of understanding both images and text, a pertinent …

abstract arxiv become capabilities cs.ai cs.cv cs.lg daily data digital focus images inference interactions language language models large language large language models llm llms model training data privacy research risks tasks training training data type vision vision-language vision-language models