Thelossesarecompared onthreedegreesofshift(easy,moderate and hard), which is controlled by the drifted distance of Gaussian clusters. Herewediscuss the architecture chosen and the implementation details. Note that the task loss / surrogate loss function is used to update the meta-loss mϕ during meta-learning. The number of transformer layers and the hidden layers in MLP are selected from{1,2}. Wecanseethatthetask loss barely affects the learnt meta loss.

For e.g., they assume pre-specified

Adversarial examples for neural network image classifiers are known to be transferable: examples optimized to be misclassified by a source classifier are often misclassified as well by classifiers with different architectures. However, targeted adversarial examples--optimized to be classified as a chosen target class--tend to be less transferable between architectures. While prior research on constructing transferable targeted attacks has focused on improving the optimization procedure, in this work we examine the role of the source classifier. Here, we show that training the source classifier to be slightly robust--that is, robust to small-magnitude adversarial examples--substantially improves the transferability of class-targeted and representation-targeted adversarial attacks, even between architectures as different as convolutional neural networks and transformers. The results we present provide insight into the nature of adversarial examples as well as the mechanisms underlying so-called robust classifiers.

The target classifier is tailored to the target data better by exploiting the low-density separation criterion.

For e.g., they assume pre-specified

For this setting, we prove that entropy minimization on unlabeled target data will avoid using the spurious feature if initialized with a decently accurate source classifier, even though the objective is non-convex and contains multiple bad local minima using the spurious features.

Adversarial examples for neural network image classifiers are known to be transferable: examples optimized to be misclassified by a source classifier are often misclassified as well by classifiers with different architectures. However, targeted adversarial examples--optimized to be classified as a chosen target class--tend to be less transferable between architectures. While prior research on constructing transferable targeted attacks has focused on improving the optimization procedure, in this work we examine the role of the source classifier. Here, we show that training the source classifier to be "slightly robust"--that is, robust to small-magnitude adversarial examples--substantially improves the transferability of class-targeted and representation-targeted adversarial attacks, even between architectures as different as convolutional neural networks and transformers. The results we present provide insight into the nature of adversarial examples as well as the mechanisms underlying so-called "robust" classifiers.