ICML 2019 accepted papers on Adversarial robustness and security
40 of the 774 papers accepted at ICML 2019 (5.2%) are on Adversarial robustness and security, compared with 2.1% at ICML 2018. They fall into 3 topics.
See ICML 2019 in the interactive map
Submitting to ICML? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Certified and empirical robustness · 26 papers
- Theoretically Principled Trade-off between Robustness and Accuracy (1,119 citations)
- Certified Adversarial Robustness via Randomized Smoothing (587 citations)
- Manifold Mixup: Better Representations by Interpolating Hidden States (470 citations)
- Using Pre-Training Can Improve Model Robustness and Uncertainty (433 citations)
- Exploring the Landscape of Spatial Robustness (250 citations)
- Improving Adversarial Robustness via Promoting Ensemble Diversity (223 citations)
- Adversarial Examples Are a Natural Consequence of Test Error in Noise (129 citations)
- Adversarial examples from computational constraints (127 citations)
- Learning with Bad Training Data via Iterative Trimmed Loss Minimization (120 citations)
- Rademacher Complexity for Adversarially Robust Generalization (116 citations)
- Interpreting Adversarially Trained Convolutional Neural Networks (69 citations)
- First-Order Adversarial Vulnerability of Neural Networks and Input Dimension (60 citations)
and 14 more in the interactive map.
Attacks on text and graphs · 8 papers
- Analyzing Federated Learning through an Adversarial Lens (382 citations)
- Imperceptible, Robust, and Targeted Adversarial Examples for Automatic Speech Recognition (229 citations)
- Adversarial Attacks on Node Embeddings via Graph Poisoning (162 citations)
- Transferable Clean-Label Poisoning Attacks on Deep Neural Nets (134 citations)
- Adversarial camera stickers: A physical camera-based attack on deep learning systems (90 citations)
- Improving Neural Language Modeling via Adversarial Training (53 citations)
- Data Poisoning Attacks in Multi-Party Learning (9 citations)
- Universal Multi-Party Poisoning Attacks (9 citations)
Black-box and transfer attacks · 6 papers
- Simple Black-box Adversarial Attacks (153 citations)
- MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement (113 citations)
- White-box vs Black-box: Bayes Optimal Strategies for Membership Inference (111 citations)
- ME-Net: Towards Effective Adversarial Robustness with Matrix Estimation (57 citations)
- Parsimonious Black-Box Adversarial Attacks via Efficient Combinatorial Optimization (54 citations)
- NATTACK: Learning the Distributions of Adversarial Examples for an Improved Black-Box Attack on Deep Neural Networks (44 citations)
