keyword

patch application operator

A patch application operator is a mathematical function used in adversarial machine learning and computer vision to superimpose an adversarial patch onto an input image. The operator accepts as inputs the patch pattern, a base image, a target spatial location, and optional transformations such as scaling, rotation, and lighting adjustments. It applies the specified transformations to the patch and overlays it onto the designated region of the image, replacing the original pixel values to produce a composite adversarial image. This operator is primarily utilized during attack generation and robustness evaluation to simulate real-world physical placements and train universal patches that remain effective across varying scenes, poses, and environmental conditions.

1 item

Adversarial Patch

Tom B. Brown, Dandelion Mané, Aurko Roy, Martín Abadi, Justin Gilmer

OrganizationsGoogle

Why you should read this

Introduces printable, scene-agnostic adversarial patches capable of overriding deep learning image classifiers in the physical world to force the prediction of an arbitrary target class.

We present a method to create universal, robust, targeted adversarial image patches in the real world. The patches are universal because they can be used to attack any scene, robust because they work under a wide variety of transformations, and targeted because they can cause a classifier to output any target class. These adversarial patches can be printed, added to any scene, photographed, and presented to image classifiers; even when the patches are small, they cause the classifiers to ignore the other items in the scene and report a chosen target class. To reproduce the results from the paper, our code is available at this https URL

Source

https://arxiv.org/pdf/1712.09665.pdf

Added

2026-09-25