New expressive losses improve adversarial robustness without sacrificing accuracy.
problem Training networks for robustness at the expense of accuracy.
method Formalizing expressivity, using convex combinations of adversarial attacks and IBP bounds.
result Trivial expressive losses yield state-of-the-art results in various settings.
New trade-off found between accuracy and adversarial robustness in regression.
problem Finding a balance between accuracy and robustness in regression models.
method Deriving a fundamental trade-off between standard and adversarial risk in regression with polynomial ridge functions.
result A necessary condition for achieving adversarial robustness without significant accuracy loss.
Despite achieving remarkable success in various domains, recent studies have uncovered the vulnerability of deep neural networks to adversarial perturbations, creating concerns on model generalizability and new threats such as prediction-evasive misclassification or stealthy reprogramming. Among different defense propo…
SCORE resolves the robustness vs accuracy trade-off by redefining robust error.
problem The inherent trade-off between robustness and accuracy in adversarial training.
method SCORE defines local equivariance as the ideal robust behavior, leading to a new robust error metric.
result SCORE reconciles robustness and accuracy, improving model performance on RobustBench.
While generalizing well over natural inputs, neural networks are vulnerable to adversarial inputs. Existing defenses against adversarial inputs have largely been detached from the real world. These defenses also come at a cost to accuracy. Fortunately, there are invariances of an object that are its salient features; w…
SAT improves adversarial training by smoothing the loss landscape through curriculum learning.
problem Adversarial training sacrifices clean accuracy for robustness and suffers from large generalization error.
method SAT uses curriculum learning to smooth the adversarial loss landscape, improving both clean and robust accuracy.
result SAT models improve clean and robust accuracy significantly compared to adversarial training and other baselines.
Adversarial training can hurt robust accuracy in small sample size scenarios.
problem Adversarial training improves test accuracy but may degrade robustness in limited data settings.
method Analyzes high-dimensional linear classification with noiseless observations, and observes perceptible attacks on image datasets.
result Adversarial training can negatively impact robust generalization in small sample size regimes.
In many real-world applications of Machine Learning it is of paramount importance not only to provide accurate predictions, but also to ensure certain levels of robustness. Adversarial Training is a training procedure aiming at providing models that are robust to worst-case perturbations around predefined points. Unfor…
Paper explores tradeoff between standard and robust accuracy for latent models.
problem Tradeoff between standard accuracy and robust accuracy in adversarial training.
method Revisits adversarial training for latent models, considering Gaussian mixture and generalized linear models.
result Low-dimensional manifold structure mitigates the tradeoff between standard and robust accuracy.
New work shows limits of certifying neural network robustness.
problem Certified training improves robustness but decreases accuracy.
method Bayes error analysis to investigate robustness limits.
result Upper bound for certified robust accuracy established.
Generative models improve adversarial robustness by adding synthetic data.
problem Improving robustness in machine learning models trained on limited data.
method Using synthetic data generated from a large dataset to augment the original training set.
result Generative models can significantly reduce the robust-accuracy gap compared to models trained with additional real data.
Deep neural networks have achieved impressive performance in many applications but their large number of parameters lead to significant computational and storage overheads. Several recent works attempt to mitigate these overheads by designing compact networks using pruning of connections. However, we observe that most …
The paper analyzes adversarial training effects on classification accuracy.
problem Understanding adversarial training's impact on standard and robust accuracy.
method Derived precise statistical analysis for binary classification problems with Gaussian data.
result Theoretical explanation of standard and robust accuracy trends for adversarial training.
New method for efficient proximal mapping of 1-path-norm in shallow networks.
problem Efficiently handling the 1-path-norm of shallow neural networks.
method Closed-form proximal operator for efficient computation and upper bound on Lipschitz constant.
result Proximal mapping allows robust training against adversarial perturbations.
Empirical adversarial risk minimization (EARM) is a widely used mathematical framework to robustly train deep neural nets (DNNs) that are resistant to adversarial attacks. However, both natural and robust accuracies, in classifying clean and adversarial images, respectively, of the trained robust models are far from sa…
Data augmentation improves robustness in adversarial training.
problem Adversarial training robust overfitting.
method Common data augmentation schemes combined with model weight averaging.
result Significant boost in robust accuracy, up to +2.93%.
SONet stabilizes ODE networks for robustness without adversarial training.
problem Improving adversarial robustness of neural networks without sacrificing natural accuracy.
method SONet uses skew-symmetric ODE blocks and DOPRI5 solver for robustness.
result SONet achieves comparable robustness to adversarial defense methods without trade-off.
While adversarial training can improve robust accuracy (against an adversary), it sometimes hurts standard accuracy (when there is no adversary). Previous work has studied this tradeoff between standard and robust accuracy, but only in the setting where no predictor performs well on both objectives in the infinite data…
Paper defends machine learning models from adversarial attacks using GLRT.
problem Adversarial attacks on machine learning models leading to misclassification.
method Generalized likelihood ratio test (GLRT) for robust classification.
result GLRT yields performance competitive with minimax approach under worst-case attacks, and better trade-off under weaker attacks.
We improve the robustness of Deep Neural Net (DNN) to adversarial attacks by using an interpolating function as the output activation. This data-dependent activation remarkably improves both the generalization and robustness of DNN. In the CIFAR10 benchmark, we raise the robust accuracy of the adversarially trained Res…
Training Deep Neural Networks that are robust to norm bounded adversarial attacks remains an elusive problem. While exact and inexact verification-based methods are generally too expensive to train large networks, it was demonstrated that bounded input intervals can be inexpensively propagated from a layer to another t…
The paper studies adversarial training for linear regression models.
problem Understanding the tradeoffs between robust and standard accuracy in adversarial training.
method Characterizes the fundamental tradeoff and specific adversarial training approach for linear regression with Gaussian features.
result Precise characterization of the standard and robust accuracy tradeoff in high-dimensional settings.
IBP-R improves verified adversarial robustness with simple, effective interval bound propagation.
problem Improving verifiability of adversarially trained networks.
method Coupling adversarial attacks with interval bound propagation for minimized verification gap.
result State-of-the-art verified robustness-accuracy trade-offs for small perturbations on CIFAR-10.
Paper provides efficient robustness certificates for neural networks.
problem Ensuring neural networks are robust against adversarial attacks.
method Two-step approach: 1) Efficient convex optimization for robustness certificates with bounded Hessian eigenvalues, 2) Curvature-based regularization during training.
result Significantly higher certified robust accuracy achieved compared to existing methods.
We demonstrate, theoretically and empirically, that adversarial robustness can significantly benefit from semisupervised learning. Theoretically, we revisit the simple Gaussian model of Schmidt et al. that shows a sample complexity gap between standard and robust classification. We prove that unlabeled data bridges thi…
Hierarchical randomized smoothing improves model robustness for complex data.
problem Certifying robustness on complex data (e.g. images, graphs) is challenging.
method Add random noise to a randomly selected subset of entities in a hierarchical manner.
result Hierarchical randomized smoothing yields stronger robustness guarantees with high accuracy.
New framework tightens certified robustness gaps in machine learning models.
problem Persistent gap between theoretical certified robustness and empirical accuracy.
method Leverages Lipschitz continuity and novel confidence intervals.
result Improves robust accuracy, compressing the gap between theory and practice.
New method shows adversarial training can be as effective as traditional training without the high cost.
problem The high cost of adversarial training limits its practical application.
method Using FGSM with random initialization and efficient training techniques.
result Adversarial training with FGSM can achieve robust accuracy comparable to PGD-based training at a lower cost.
RayS attack improves hard-label adversarial attacks by reducing query complexity and identifying false robust models.
problem Challenges in hard-label adversarial attacks, especially in terms of effectiveness and efficiency.
method Reformulates continuous problem into discrete problem without gradient estimation and uses a fast check step to eliminate unnecessary searches.
result Significantly reduces the number of queries needed for hard-label attacks and identifies false robust models.
Paper tackles catastrophic overfitting in single-step adversarial training.
problem Catastrophic overfitting leads to sudden drop in robust accuracy.
method Proposes a method to prevent overfitting by using all adversarial examples.
result Demonstrates prevention of catastrophic overfitting and improves robustness.
Survey of weak form's role in equation learning, parameter estimation, and coarse graining.
problem Noise robustness, accuracy, and computational efficiency in weak form applications.
method Survey and recent developments in weak form versions of equation learning, parameter estimation, and coarse graining.
result Surprising noise robustness, accuracy, and computational efficiency in weak form applications.
This work benchmarks and theorizes robust NAS under adversarial training.
problem Lack of benchmark evaluations and theoretical guarantees for robust NAS architectures under adversarial training.
method Released a comprehensive data set and established a generalization theory using the neural tangent kernel.
result Established a generalization theory for robust NAS architectures under adversarial training.
New method calibrates ambiguity sets for robust decision-making under contamination.
problem Minimizing worst-case expected loss over distributional shifts in out-of-sample environments.
method Bulk-calibrated credal ambiguity sets that learn a high-mass bulk set from data and bound tail contributions.
result Closed-form, finite robust objective and tractable optimization for various losses and geometries.
Improves authorship attribution accuracy and robustness against adversarial attacks.
problem Vulnerability of authorship attribution approaches to adversarial attacks.
method Formulated a relational adversary threat model and presented a novel learning framework, normalize-and-predict (N&P).
result Improves accuracy on adversarial inputs by up to 70% and robust accuracy by 45%.
New methods show robustness and accuracy can coexist.
problem Inevitability of robustness-accuracy tradeoff in deep learning.
method Prove robustness and accuracy achievable through locally Lipschitz functions; explore combining dropout with robust training methods.
result Achieving robustness and accuracy requires methods imposing local Lipschitzness and deep learning generalization techniques.
Paper analyzes robustness of non-Lipschitz networks, proving powerful adversarial attacks but offering solutions.
problem Adversarial attacks on deep networks, especially non-Lipschitz networks.
method Developed an attack model that abstracts the challenge of adversarial robustness, proving the power of such attacks and offering solutions.
result Proves powerful adversarial attacks on non-Lipschitz networks but offers solutions with abstention.
Neural networks are vulnerable to small adversarial perturbations. Existing literature largely focused on understanding and mitigating the vulnerability of learned models. In this paper, we demonstrate an intriguing phenomenon about the most popular robust training method in the literature, adversarial training: Advers…
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.
Voting ensemble of robust models improves robustness.
problem Improving robustness of defensive models against adversarial attacks.
method Hard-label voting ensemble of pretrained robust models.
result Voting ensemble can boost robust error over individual models.
Improving the accuracy and robustness of deep neural nets (DNNs) and adapting them to small training data are primary tasks in deep learning research. In this paper, we replace the output activation function of DNNs, typically the data-agnostic softmax function, with a graph Laplacian-based high dimensional interpolati…
Paper proposes a method to train robust neural networks without labeled data.
problem Training robust neural networks without class labels.
method Adversarial contrastive learning framework using unlabeled data.
result Robust Contrastive Learning (RoCL) achieves comparable robust accuracy to supervised methods and significantly improved robustness.
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.
Unified smoothing for robust classification improves accuracy.
problem Improving robustness of classifiers against adversarial attacks.
method Learned smoothed densities and randomized smoothing.
result Provable robust accuracies higher than state-of-the-art defenses.
High-dimensional neural network manifolds misalign with human perception, causing adversarial examples.
problem Adversarial attacks fool neural networks, but their origin is unclear.
method Defined and analyzed a network's perceptual manifold (PM) for a class concept.
result Neural network PMs have orders of magnitude higher dimensions than natural human concepts, suggesting exponential misalignment.
LOTOS improves ensemble robustness by promoting orthogonal transformations.
problem Transferability of adversarial examples threatens robustness of classification models.
method LOTOS promotes orthogonality among sub-spaces of transformations in ensemble models.
result LOTOS increases robust accuracy of ensembles by 6 percentage points against black-box attacks.
Paper develops a robust classifier for Gaussian mixture models under sparse adversarial perturbations.
problem Classifying data under sparse adversarial perturbations for Gaussian mixture models.
method Develops FilTrun algorithm with filtration and truncation modules.
result Characterizes optimal robust classifier and robust classification error.
Validates neural networks inputs to protect against adversarial examples.
problem Ensuring neural networks robustness against adversarial attacks.
method Runtime local robustness verification based on normal distribution of robustness radii.
result Improves neural network accuracy and protects against adversarial examples.
Deep networks are well-known to be fragile to adversarial attacks. We conduct an empirical analysis of deep representations under the state-of-the-art attack method called PGD, and find that the attack causes the internal representation to shift closer to the "false" class. Motivated by this observation, we propose to …