Imperceptible Adversarial Examples by Spatial Chroma-Shift
This addresses the challenge of making adversarial attacks more stealthy for security and robustness testing in computer vision, though it is incremental as it builds on existing spatial transformation methods.
The paper tackled the problem of creating adversarial examples that are imperceptible to humans by proposing a method that modifies only the color components of images, leveraging human vision's lower sensitivity to chrominance distortions. The result showed competitive fooling rates on datasets like CIFAR-10 and NIPS2017, with improved perceptual quality metrics and human studies confirming the examples are often indistinguishable from originals.
Deep Neural Networks have been shown to be vulnerable to various kinds of adversarial perturbations. In addition to widely studied additive noise based perturbations, adversarial examples can also be created by applying a per pixel spatial drift on input images. While spatial transformation based adversarial examples look more natural to human observers due to absence of additive noise, they still possess visible distortions caused by spatial transformations. Since the human vision is more sensitive to the distortions in the luminance compared to those in chrominance channels, which is one of the main ideas behind the lossy visual multimedia compression standards, we propose a spatial transformation based perturbation method to create adversarial examples by only modifying the color components of an input image. While having competitive fooling rates on CIFAR-10 and NIPS2017 Adversarial Learning Challenge datasets, examples created with the proposed method have better scores with regards to various perceptual quality metrics. Human visual perception studies validate that the examples are more natural looking and often indistinguishable from their original counterparts.