# Budget-Aware Adaptive Adversarial Patches for Black-Box Object Detection

Source: [Budget-Aware Adaptive Adversarial Patches for Black-Box Object Detection](<https://www.sota2.com/research/paper/6a337b69206542251d7c4f11>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-19T05:00:05.274Z

Authors: Pedram MohajerAnsari, Amir Salarpour, David Fernandez, Mert D. Pes\\'e

Published: 2026-06-16T13:38:15.000Z

## Abstract

Adversarial patches pose a practical threat to modern object detectors. Prior work shows vulnerability, but three gaps limit actionable insight: (i) few \\emph{score-based black-box} attacks \\emph{jointly} optimize patch \\emph{location, texture, and size} under tight query budgets; (ii) success is rarely tied to the patch's \\emph{visual footprint}; and (iii) evaluations often conflate EOT robustness with plain-view suppression. We present \\method{}, a query-efficient, budget-adaptive black-box attack that couples a lightweight \\emph{Contextual Thompson-Sampling} placer with NES-style pixel updates, growing the patch only when progress stalls. Reporting is anchored by a \\emph{strict plain-image} suppression test; EOT is audited but never used as a substitute for success, and optional appearance/printability weights expose strength--visibility trade-offs. Across YOLOv5, Faster R-CNN, and YOLOS, \\method{} achieves strong suppression on CNN-based detectors and substantial suppression on the transformer-based detector, using compact patches and exposing clear query--footprint trade-offs relative to fixed-size and heuristic baselines. A print--capture pilot further shows transfer across unseen physical objects and viewpoints.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Object Detection on PASCAL VOC 2007](<https://www.sota2.com/research/sota/object-detection-on-pascal-voc-2007.md>): Object Detection; PASCAL VOC 2007; mAP: not recorded.

- [Image Classification on Classification Dataset](<https://www.sota2.com/research/sota/image-classification-on-classification-dataset.md>): Image Classification; Classification Dataset; Top-1 Accuracy: not recorded.

- [Object Detection on Person Detection Dataset](<https://www.sota2.com/research/sota/object-detection-on-person-detection-dataset.md>): Object Detection; Person Detection Dataset; AP: not recorded.

- [Object Detection on Stop-sign Detection Dataset](<https://www.sota2.com/research/sota/object-detection-on-stop-sign-detection-dataset.md>): Object Detection; Stop-sign Detection Dataset; Stop-sign Miss Rate: not recorded.

- [Object Detection on Authors' Evaluation Protocol Dataset](<https://www.sota2.com/research/sota/object-detection-on-authors-evaluation-protocol-dataset.md>): Object Detection; Authors' Evaluation Protocol Dataset; AP: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.18318>)

[PDF](<https://arxiv.org/pdf/2606.18318.pdf>)
