Structured Adversarial Camouflage via Voronoi Diagrams

2026-06-16

AI Summary

This work addresses the high computational cost and visual conspicuousness of conventional pixel-level adversarial patches, which hinder their deployment in safety-critical systems. The authors propose a parameter-efficient adversarial camouflage method that optimizes seed point locations under a constrained printable color palette and leverages Voronoi diagrams to generate structured camouflage patterns. This approach effectively disrupts object detectors without requiring additional regularization, jointly optimizing pattern structure and palette design to balance visual plausibility with attack efficacy. The method demonstrates strong black-box transferability across models and domains. Experiments on the 3DPeople dataset show a significant reduction in COCO AP@[.5:.95] for YOLOv9–v12 detectors, with attack performance generalizing to out-of-domain backgrounds, though it exhibits sensitivity to variations in the color palette.

Abstract

Pixel-wise adversarial patches are computationally heavy and often visually detectable, limiting utility in security-critical systems. We present adversarial Voronoi camouflage that optimizes only seed-point locations under fixed, printable palettes using a soft assignment, producing structured, splinter camouflage-like patterns without additional regularization. Evaluated on person detection with COCO-style AP@[.5:.95], naive placement (Inria -> COCO) performs comparably bad, while garment-level application via segmentation mask (3DPeople) results in a significant AP drop. The attack transfers to out-of-domain backgrounds and across detector families (YOLOv9/10/11/12), indicating robustness in black-box settings. Repainting with different palettes largely nullifies the effect, and single-color tweaks show limited tolerance (<=0.17), highlighting a structure-palette coupling. The parameter-efficient, palette-constrained design improves visual plausibility while degrading real-time detector performance. Physical validation and color calibration are left for future work. Code: https://github.com/JensBayer/Voronoi This paper was originally presented at the International Conference on Military Communication and Information Systems (ICMCIS), organized by the Information Systems Technology (IST) Scientific and Technical Committee, IST-224-RSY - the ICMCIS, held in Bath, United Kingdom, 12-13 May 2026.

Problem

• adversarial camouflage
• pixel-wise adversarial patches
• visual detectability
• computational efficiency
• security-critical systems

Innovation

• adversarial camouflage
• Voronoi diagrams
• palette-constrained optimization
• structured perturbations
• black-box transferability