GeoCloak: An Agentic Framework for Semantics-Preserving Geolocation Privacy via Peripheral Clue Manipulation
GeoCloak cuts LVLM city-level geolocation accuracy from 41.2% to 9.9% while preserving image meaning.
GeoCloak is an agentic framework that protects image geolocation privacy against large vision-language models without relying only on visible adversarial noise. It identifies location-relevant but visually peripheral clues, applies targeted semantic edits, and adds configured perturbation under utility-aware feedback. Across four datasets spanning 108 countries and 576 cities, tested against five LVLMs and three inference methods, it cut city-level accuracy from 41.2% to 9.9%, about half the best baseline, while preserving image utility better.
- GeoCloak targets peripheral visual clues that dominate geolocation inference.
- It combines semantic edits with residual adversarial perturbation.
- City-level accuracy fell from 41.2% to 9.9% across five LVLMs.
- Evaluation covered four datasets, 108 countries, and 576 cities.
Full article206 words · extracted from arxiv.org · click to collapse
The rapid advancement of large vision language models (LVLMs) has made geolocation inference from images highly accurate and accessible. This poses serious privacy risks: a passive adversary with black-box LVLM access can infer a user's location from a single photo with no metadata. Existing protection methods rely almost exclusively on adversarial perturbation, which introduces perceptible visual artifacts and utility degradation. We identify the visual-geolocation asymmetry: geolocation-decisive clues, such as street signs, tend to be visually peripheral but dominate location inference. Exploiting this asymmetry, we propose GeoCloak, an agentic framework for semantics-preserving geolocation privacy protection. It acts as a clue-aware planner that analyzes each image to identify geolocation-relevant clues, dispatches targeted semantic edits to directly suppress peripheral clues, and autonomously configures adversarial perturbation to further suppress residual geolocation features, all under iterative utility-aware feedback. We evaluate GeoCloak on four diverse datasets spanning 108 countries and 576 cities against five LVLMs and three inference methods. GeoCloak reduces overall city-level accuracy from 41.2% to 9.9%, approximately half of the best-performing baseline, while achieving substantially superior utility preservation both qualitatively and quantitatively. We hope this work takes a meaningful step toward practical, semantics-preserving geolocation privacy protection, giving users greater control over the location information inadvertently embedded in their shared images.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2610.20055