Rethinking Visual Privacy: A Compositional Privacy Risk Framework for Severity Assessment with VLMs
This work addresses visual privacy for AI systems by moving beyond binary classification to a compositional approach, offering a novel framework and dataset for more nuanced risk assessment.
The paper tackles the problem of visual privacy assessment by proposing that privacy is compositional, where benign attributes combine to cause severe violations, and introduces a framework (CPRT) with graded severity levels and a scoring function, achieving a deployable 8B SFT model that matches frontier-level performance.
Existing visual privacy benchmarks largely treat privacy as a binary property, labeling images as private or non-private based on visible sensitive content. We argue that privacy is fundamentally compositional. Attributes that are benign in isolation may combine to produce severe privacy violations. We introduce the Compositional Privacy Risk Taxonomy (CPRT), a regulation-aware framework that organizes visual attributes according to standalone identifiability and compositional harm potential. CPRT defines four graded severity levels and is paired with an interpretable scoring function that assigns continuous privacy severity scores. We further construct a taxonomy-aligned dataset of 6.7K images and derive ground-truth compositional risk scores. By evaluating frontier and open-weight VLMs we find that frontier models align well with compositional severity when provided structured guidance, but systematically underestimate composition-driven risks. Smaller models struggle to internalize graded privacy reasoning. To bridge this gap, we introduce a deployable 8B supervised fine-tuned (SFT) model that closely matches frontier-level performance on compositional privacy assessment.