Unrestricted Adversarial Examples via Semantic Manipulation
This addresses the vulnerability of machine learning models to adversarial attacks by creating more realistic and harder-to-defend examples, which is an incremental improvement over existing methods.
The paper tackles the problem of adversarial examples by introducing unrestricted perturbations that manipulate color and texture to generate photorealistic attacks, showing effectiveness against defenses like JPEG compression and adversarially trained models on datasets such as ImageNet and MSCOCO.
Machine learning models, especially deep neural networks (DNNs), have been shown to be vulnerable against adversarial examples which are carefully crafted samples with a small magnitude of the perturbation. Such adversarial perturbations are usually restricted by bounding their $\mathcal{L}_p$ norm such that they are imperceptible, and thus many current defenses can exploit this property to reduce their adversarial impact. In this paper, we instead introduce "unrestricted" perturbations that manipulate semantically meaningful image-based visual descriptors - color and texture - in order to generate effective and photorealistic adversarial examples. We show that these semantically aware perturbations are effective against JPEG compression, feature squeezing and adversarially trained model. We also show that the proposed methods can effectively be applied to both image classification and image captioning tasks on complex datasets such as ImageNet and MSCOCO. In addition, we conduct comprehensive user studies to show that our generated semantic adversarial examples are photorealistic to humans despite large magnitude perturbations when compared to other attacks.