Free adversarial training reduces the generalization gap compared to vanilla method.
problem Improving generalization in adversarial training.
method Analysis of algorithmic stability in free adversarial training.
result Free adversarial training shows a lower generalization gap.
Adversarial training improves graph autoencoder generalization.
problem Improving graph autoencoder generalization.
method Formulated L2 and L1 adversarial training for graph autoencoders and variational graph autoencoders.
result Adversarial training boosts graph autoencoder and variational graph autoencoder generalization.
By injecting adversarial examples into training data, adversarial training is promising for improving the robustness of deep learning models. However, most existing adversarial training approaches are based on a specific type of adversarial attack. It may not provide sufficiently representative samples from the adversa…
We study two important concepts in adversarial deep learning---adversarial training and generative adversarial network (GAN). Adversarial training is the technique used to improve the robustness of discriminator by combining adversarial attacker and discriminator in the training phase. GAN is commonly used for image ge…
New approach improves adversarial robustness without sacrificing natural generalization.
problem Balancing adversarial robustness and natural generalization in machine learning.
method Friendly adversarial training (FAT) using early-stopped PGD to find least adversarial data.
result Early-stopped PGD achieves adversarial robustness without compromising natural generalization.
The paper studies the asymptotic behavior of adversarial training under ℓ∞-perturbation.
problem Theoretical guarantees for sparsity-recovery in adversarial training.
method Investigation of the asymptotic distribution of the adversarial training estimator in generalized linear models.
result The asymptotic distribution of the adversarial training estimator under ℓ∞-perturbation could have a positive probability mass at 0 when the true parameter is 0. CCAT improves model robustness to various adversarial attacks.
problem Robustness to adversarial attacks does not generalize to unseen threat models.
method CCAT biases models towards low confidence predictions on adversarial examples.
result CCAT increases robustness against multiple adversarial attack norms and types.
Adversarial training, in which a network is trained on adversarial examples, is one of the few defenses against adversarial attacks that withstands strong attacks. Unfortunately, the high cost of generating strong adversarial examples makes standard adversarial training impractical on large-scale problems like ImageNet…
This paper improves adversarial robustness of deep learning models.
problem Vulnerability of machine learning models to adversarial perturbations.
method Analyzes adversarial training for linear regression and neural networks, incorporating L1 penalty.
result Incorporating L1 penalty leads to consistent adversarially robust estimation in high-dimensional settings.
A new method reduces adversarial training time without overfitting.
problem Catastrophic overfitting in single-step adversarial training.
method FGSMPR: FGSM with PGD Regularization.
result Reduces the gap to multi-step adversarial training.
Interpolated adversarial training improves robustness without sacrificing accuracy.
problem Improving adversarial robustness without degrading generalization performance.
method Interpolated adversarial training using interpolation-based methods.
result Retains adversarial robustness while achieving a standard test error of 6.45%.
Simple regularization methods mimic adversarial training's robustness.
problem Expensive adversarial training for robustness.
method Label smoothing and logit squeezing.
result Achieves strong adversarial robustness without adversarial examples.
SPAT improves adversarial robustness by preserving semantics in adversarial training.
problem Adversarial examples often have different semantics than original data, introducing unintended biases.
method Semantics-preserving adversarial training (SPAT) that encourages pixel perturbation shared among all classes.
result SPAT improves adversarial robustness and achieves state-of-the-art results in CIFAR-10 and CIFAR-100.
More training data can hurt the generalization of adversarially robust models.
problem The challenge of balancing adversarial robustness and generalization in machine learning models.
method Investigation of three regimes based on adversary strength and empirical studies on various models.
result More training data can hurt the generalization of adversarially robust models in different regimes.
This paper quantifies privacy-robustness and generalization-robustness trade-offs in adversarial training.
problem Privacy and generalization issues in adversarial training.
method Defines robustified intensity and empirical robustified intensity to measure robustness, proving differential privacy and generalization bounds.
result Proves adversarial training is (ε,δ)-differentially private and provides generalization bounds. Meta-CoTGAN improves adversarial text generation by preventing mode collapse.
problem Mode collapse in adversarial text generation.
method Meta-Cooperative Training Paradigm with a language model.
result Meta-CoTGAN effectively slows down mode collapse and improves generation quality and diversity.
Localized adversarial training improves image classifiers' robustness.
problem State-of-the-art image classifiers fail on carefully manipulated adversarial images.
method Developed a localized adversarial attack and used it to train a robust classifier.
result Localized adversarial training increases robustness against adversarial inputs.
BPN defends against adversarial attacks by generating beneficial perturbations.
problem Adversarial attacks cause deep neural networks to misclassify clean inputs.
method BPN generates beneficial perturbations during training to neutralize future adversarial attacks.
result BPN is robust to adversarial examples and more efficient than classical adversarial training.
Adversarial training makes models more vulnerable to privacy attacks.
problem Privacy attacks on robust models become feasible.
method Demonstrated through model inversion attacks on robustly trained models.
result Privacy attacks on robust models are now feasible.
Injecting adversarial examples during training, known as adversarial training, can improve robustness against one-step attacks, but not for unknown iterative attacks. To address this challenge, we first show iteratively generated adversarial images easily transfer between networks trained with the same strategy. Inspir…
Paper explores why overfitted DNNs in adversarial training can generalize.
problem Understanding why overfitted DNNs in adversarial training can generalize despite poor robust generalization.
method An approximation viewpoint to analyze the robust overfitting of over-parameterized DNNs.
result Existence of infinitely many overfitted DNNs that achieve good robust generalization under certain conditions.
Adversarial training makes logistic regression weight loss landscapes sharper.
problem Understanding why adversarial training sharpens the weight loss landscape in logistic regression.
method Theoretical analysis of linear logistic regression model with L2 norm constraints, and experiments on ResNet18.
result Adversarial training sharpens the weight loss landscape in linear logistic regression models.
Adversarial learning of probabilistic models has recently emerged as a promising alternative to maximum likelihood. Implicit models such as generative adversarial networks (GAN) often generate better samples compared to explicit models trained by maximum likelihood. Yet, GANs sidestep the characterization of an explici…
As adversarial attacks pose a serious threat to the security of AI system in practice, such attacks have been extensively studied in the context of computer vision applications. However, few attentions have been paid to the adversarial research on automatic path finding. In this paper, we show dominant adversarial exam…
Adversarial training achieves optimal test error for shallow networks.
problem Achieving optimal adversarial test error for general data distributions.
method Applying new Rademacher complexity bounds and properties of optimal adversarial predictors.
result Adversarial training can achieve optimal adversarial test error for general data distributions.
The paper analyzes how generated data improves adversarial training in high-dimensional regression.
problem Improving adversarial training in high-dimensional regression.
method Theoretical analysis of a two-stage training approach with generated data and pseudo-labels.
result Two-stage adversarial training achieves better performance than ridgeless training in high-dimensional linear regression.
New method improves DNN robustness against adversarial attacks.
problem Adversarial vulnerability of neural networks.
method Enforcing linearity in DNN in transformed input and feature space, and augmenting with Local Lipschitz regularizer.
result Significantly improved robustness of DNN models on various datasets.
CAT improves robustness of neural networks by customizing perturbation levels.
problem Poor generalization of adversarial training on clean and perturbed data.
method CAT adapts perturbation level and label for each sample.
result CAT achieves better clean and robust accuracy than previous methods.
Paper analyzes adversarial training's performance in binary classification.
problem Understanding the generalization performance of adversarial training.
method Derives precise theoretical predictions for adversarial training performance.
result Provides exact asymptotics for test errors of adversarial training.
Adversarial samples are perturbed inputs crafted to mislead the machine learning systems. A training mechanism, called adversarial training, which presents adversarial samples along with clean samples has been introduced to learn robust models. In order to scale adversarial training for large datasets, these perturbati…
Improved neural network robustness with instance-specific perturbation margins.
problem Adversarial training fails to generalize well to unperturbed test set.
method Instance adaptive adversarial training with sample-specific perturbation margins.
result Test accuracy improves with a marginal drop in robustness.
Adversarial training provides a principled approach for training robust neural networks. From an optimization perspective, adversarial training is essentially solving a bilevel optimization problem. The leader problem is trying to learn a robust classifier, while the follower problem is trying to generate adversarial s…
AWP improves robustness by flattening weight loss landscape.
problem Improving robustness of deep neural networks against adversarial examples.
method Explicitly regularizes the flatness of weight loss landscape through adversarial weight perturbation.
result AWP forms a double-perturbation mechanism in adversarial training, leading to flatter weight loss landscape.
Generative adversarial networks (GAN) have been effective for learning generative models for real-world data. However, existing GANs (GAN and its variants) tend to suffer from training problems such as instability and mode collapse. In this paper, we propose a novel GAN framework called evolutionary generative adversar…
We propose a novel technique to make neural network robust to adversarial examples using a generative adversarial network. We alternately train both classifier and generator networks. The generator network generates an adversarial perturbation that can easily fool the classifier network by using a gradient of each imag…
Proposes SWA for adversarial training to improve model robustness.
problem Overfitting in adversarial training due to insufficient sample complexity.
method Adversarial training with Stochastic Weight Average (SWA).
result Improves model robustness on CIFAR-10, CIFAR-100, and SVHN datasets.
Deep neural networks (DNNs) have been found to be vulnerable to adversarial examples resulting from adding small-magnitude perturbations to inputs. Such adversarial examples can mislead DNNs to produce adversary-selected results. Different attack strategies have been proposed to generate adversarial examples, but how t…
Study analyzes adversarial training dynamics without data distribution assumptions.
problem Understanding training dynamics of adversarial training without data distribution assumptions.
method Mean field theory approach to analyze adversarial training in random deep neural networks.
result Upper bounds of adversarial loss derived empirically and theoretically.
New adversarial training methods generate multiplicative perturbations for robust DNN training.
problem Training Deep Neural Networks with adversarial examples to improve robustness.
method Proposes xAT and xVAT, generating multiplicative perturbations for robust training.
result xAT and xVAT match or outperform state-of-the-art classification accuracies and are faster.
Adversarial training leads to clean data generalization with significant robust overfitting gap.
problem Significant robust generalization gap in adversarial training.
method Two theoretical views: representation complexity and training dynamics.
result ReLU nets with O(ND) extra parameters can achieve CGRO. Deep neural networks with adversarial training achieve sup-norm convergence for nonparametric regression.
problem Achieving sup-norm convergence for deep neural network estimators in nonparametric regression.
method Developed an adversarial training scheme to address the sup-norm convergence issue.
result Deep neural network estimators achieve optimal sup-norm convergence with the proposed adversarial training.
Theoretical study shows adversarial training improves robustness in deep learning models.
problem Ensuring robustness in pre-trained deep learning models.
method Theoretical analysis of adversarial training and feature purification in two-layer neural networks.
result Adversarial training leads to feature purification, making models more robust to attacks.
Paper tackles robustness in adversarial noise with a meta-optimizer.
problem Sensitivity to adversarial noise hinders machine learning deployment.
method Meta-optimizer learns to robustly optimize models using adversarial examples.
result Meta-optimizer transfers adversarial knowledge to new models without generating new examples.
TEAM uses Taylor expansion to generate adversarial examples.
problem Vulnerability of deep neural networks to adversarial examples.
method Approximates DNN output using Taylor expansion and optimizes with Lagrange multiplier method.
result Improves robustness of DNNs through adversarial training.
In recent years, deep neural networks have demonstrated outstanding performance in many machine learning tasks. However, researchers have discovered that these state-of-the-art models are vulnerable to adversarial examples: legitimate examples added by small perturbations which are unnoticeable to human eyes. Adversari…
This paper explores adversarial robustness of flow-based generative models.
problem Robustness of flow-based generative models to adversarial attacks.
method Theoretical and empirical analysis of adversarial robustness for simple and complex flow-based models.
result Flow-based generative models are highly sensitive to adversarial attacks, but robustness can be significantly improved using hybrid adversarial training.
New findings suggest adversarial training does not flatten loss landscapes, challenging current intuition.
problem Understanding and improving generalization in deep learning.
method Loss surface visualization with filter normalization technique.
result Adversarial training does not result in flatter loss landscapes, challenging current intuition.
Adversarial training can lead to overfitting without compromising robustness.
problem Explaining benign overfitting in adversarially robust linear classification.
method Theoretical analysis and numerical experiments on adversarial training.
result Adversarially trained linear classifiers can achieve near-optimal risks despite overfitting noisy data.