arXiv research

Our work investigates how to identify privacy violations in models using finite adversaries.

Adversarial examples pose a threat to deep neural network models in a variety of scenarios, from settings where the adversary has complete knowledge of the model and to the opposite "black box" setting. Black box attacks are particularly threatening as the adversary only needs access to the input and output of the mode…

Paper shows MoM is optimal under adversarial contamination for certain distributions.

We study the question of learning an adversarially robust predictor. We show that any hypothesis class $\mathcal{H}$ with finite VC dimension is robustly PAC learnable with an improper learning rule. The requirement of being improper is necessary as we exhibit examples of hypothesis classes $\mathcal{H}$ with finite VC…

Study on how adversarial training affects neural network kernels and robustness.

Bayesian algorithms improve online learning with adversaries over infinite action spaces.

A federated learning algorithm tackles linear bandits with adversarial actions, achieving optimal regret bounds.

The study sets lower bounds on MMSE for inferring sensitive features from noisy data.

Study on proper learning under relaxed worst-case robust loss for VC classes.

Generates positive examples from noisy data streams.

Study robustness of split conformal prediction under adversarial attacks.

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…

Study adversarial attacks on cost-sensitive classifiers.

New method improves robustness in partially observable domains by training against latent distribution shifts.

Study robust regression learning under adversarial attacks.

Study reveals how high-dimensional models are vulnerable to consistent adversarial attacks.

Adversarial training improves linear regression solutions, offering robustness against small perturbations.

Optimizes quantile and semi-adversarial regret with novel root-logarithmic regularizers.

Note that this paper is superceded by "Black-Box Adversarial Attacks with Limited Queries and Information." Current neural network-based image classifiers are susceptible to adversarial examples, even in the black-box setting, where the attacker is limited to query access without access to gradients. Previous methods -…

Adversarial online multi-task RL with task separation.

Study statistical guarantees for DRO with OT and OT-regularized divergences.

New method approximates neural network training for robustness.

Differentiable language model attacks improve adversarial examples for categorical sequence classifiers.

Study on robustness of learning-based novelty detection methods under adversarial attacks.

We show in this note that the Sobolev Discrepancy introduced in Mroueh et al in the context of generative adversarial networks, is actually the weighted negative Sobolev norm $||.||_{\dot{H}^{-1}(ν_q)}$, that is known to linearize the Wasserstein $W_2$ distance and plays a fundamental role in the dynamic formulation of…

Adversarially robust machine learning has received much recent attention. However, prior attacks and defenses for non-parametric classifiers have been developed in an ad-hoc or classifier-specific basis. In this work, we take a holistic look at adversarial examples for non-parametric classifiers, including nearest neig…

New algorithms for regression with adversarial responses on various metric spaces.

Study shows multi-source learning is more resilient to adversarial corruption than single-source learning.

Paper analyzes adversarial attacks and defenses using game theory.

We describe an adversarial learning approach to constrain convolutional neural network training for image registration, replacing heuristic smoothness measures of displacement fields often used in these tasks. Using minimally-invasive prostate cancer intervention as an example application, we demonstrate the feasibilit…

New algorithm resists contamination in high-dimensional regression with optimal performance.

Study improves distributed linear estimation under adversarial conditions.

Solves expert prediction problem for 4 experts in finite time horizon.

The paper analyzes adversarial robustness of Gaussian processes.

Algorithm learns without knowing distribution, reducing error.

Framework detects and mitigates data-poisoning attacks in causal effect estimation.

New model selects robustly in adversarial reinforcement learning with unknown corruption.

Framework learns robust control policies from expert demonstrations.

Partial monitoring is a generalization of the well-known multi-armed bandit framework where the loss is not directly observed by the learner. We complete the classification of finite adversarial partial monitoring to include all games, solving an open problem posed by Bartok et al. [2014]. Along the way we simplify and…

The paper explores robust classifiers for imbalanced Gaussian data.

This paper investigates stochastic and adversarial combinatorial multi-armed bandit problems. In the stochastic setting under semi-bandit feedback, we derive a problem-specific regret lower bound, and discuss its scaling with the dimension of the decision space. We propose ESCB, an algorithm that efficiently exploits t…

Safe imitation learning with a safety layer for flexible training.

New method verifies neural network robustness globally using generative models.

Adversarial meta-learning computes Gamma-minimax estimators for vague prior knowledge.

Deep neural networks suffer from over-fitting and catastrophic forgetting when trained with small data. One natural remedy for this problem is data augmentation, which has been recently shown to be effective. However, previous works either assume that intra-class variances can always be generalized to new classes, or e…

We introduce new families of Integral Probability Metrics (IPM) for training Generative Adversarial Networks (GAN). Our IPMs are based on matching statistics of distributions embedded in a finite dimensional feature space. Mean and covariance feature matching IPMs allow for stable training of GANs, which we will call M…

New insights into learning from distributional adversaries and private data.

We study the problem of learning Markov decision processes with finite state and action spaces when the transition probability distributions and loss functions are chosen adversarially and are allowed to change with time. We introduce an algorithm whose regret with respect to any policy in a comparison class grows as t…