GAP++: Learning to generate target-conditioned adversarial examples
This work addresses the challenge of efficient and effective adversarial attacks for machine learning models, offering a more general-purpose approach compared to previous single-target methods.
The paper tackles the problem of generating adversarial examples by proposing a target-conditioned framework that learns perturbations based on both input images and target labels, achieving superior performance and high fooling rates with small perturbation norms on MNIST and CIFAR10 datasets.
Adversarial examples are perturbed inputs which can cause a serious threat for machine learning models. Finding these perturbations is such a hard task that we can only use the iterative methods to traverse. For computational efficiency, recent works use adversarial generative networks to model the distribution of both the universal or image-dependent perturbations directly. However, these methods generate perturbations only rely on input images. In this work, we propose a more general-purpose framework which infers target-conditioned perturbations dependent on both input image and target label. Different from previous single-target attack models, our model can conduct target-conditioned attacks by learning the relations of attack target and the semantics in image. Using extensive experiments on the datasets of MNIST and CIFAR10, we show that our method achieves superior performance with single target attack models and obtains high fooling rates with small perturbation norms.