The Pervasive Blind Spot: Benchmarking VLM Inference Risks on Everyday Personal Videos

TL;DR

This paper benchmarks the inference risks of vision-language models on personal videos, revealing their superhuman capabilities, the influence of video factors on privacy risks, and the unreliability of model explanations.

Contribution

It introduces a new dataset of personal videos and systematically evaluates VLM inference risks, highlighting their privacy implications and limitations in explainability.

Findings

VLMs outperform humans in inferring sensitive attributes.

Inference risk depends on video and prompting factors.

Model explanations are often unreliable and misleading.

Abstract

The proliferation of Vision-Language Models (VLMs) introduces profound privacy risks from personal videos. This paper addresses the critical yet unexplored inferential privacy threat, the risk of inferring sensitive personal attributes over the data. To address this gap, we crowdsourced a dataset of 508 everyday personal videos from 58 individuals. We then conducted a benchmark study evaluating VLM inference capabilities against human performance. Our findings reveal three critical insights: (1) VLMs possess superhuman inferential capabilities, significantly outperforming human evaluators, leveraging a shift from object recognition to behavioral inference from temporal streams. (2) Inferential risk is strongly correlated with factors such as video characteristics and prompting strategies. (3) VLM-driven explanation towards the inference is unreliable, as we revealed a disconnect between the model-generated explanations and evidential impact, identifying ubiquitous objects as misleading confounders.