NeurIPS 2021 accepted papers on Adversarial robustness and security
91 of the 2,508 papers accepted at NeurIPS 2021 (3.6%) are on Adversarial robustness and security, compared with 3.8% at NeurIPS 2020. They fall into 4 topics.
See NeurIPS 2021 in the interactive map
Submitting to NeurIPS? 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 · 55 papers
- Are Transformers more robust than CNNs? (92 citations)
- Do Wider Neural Networks Really Help Adversarial Robustness? (35 citations)
- Exploring Architectural Ingredients of Adversarially Robust Deep Neural Networks (32 citations)
- Towards Efficient and Effective Adversarial Training (30 citations)
- Adversarial Attack Generation Empowered by Min-Max Optimization (21 citations)
- Probabilistic Margins for Instance Reweighting in Adversarial Training (21 citations)
- Distilling Robust and Non-Robust Features in Adversarial Examples by Information Bottleneck (17 citations)
- AugMax: Adversarial Composition of Random Augmentations for Robust Training (16 citations)
- Class-Disentanglement and Applications in Adversarial Detection and Defense (16 citations)
- Adversarial Robustness of Streaming Algorithms through Importance Sampling (15 citations)
- Robust Deep Reinforcement Learning through Adversarial Loss (15 citations)
- Fast Minimum-norm Adversarial Attacks through Adaptive Norm Constraints (14 citations)
and 43 more in the interactive map.
Attacks on text and graphs · 20 papers
- Adversarial GLUE: A Multi-Task Benchmark for Robustness Evaluation of Language Models (72 citations)
- Adversarial Examples Make Strong Poisons (34 citations)
- Variational Model Inversion Attacks (30 citations)
- How Should Pre-Trained Language Models Be Fine-Tuned Towards Adversarial Robustness? (27 citations)
- Learning Transferable Adversarial Perturbations (27 citations)
- Unadversarial Examples: Designing Objects for Robust Vision (25 citations)
- Predify: Augmenting deep neural networks with brain-inspired predictive coding dynamics (24 citations)
- Not All Low-Pass Filters are Robust in Graph Convolutional Networks (19 citations)
- Gradient-Free Adversarial Training Against Image Corruption for Learning-based Steering (9 citations)
- Graph Robustness Benchmark: Benchmarking the Adversarial Robustness of Graph Machine Learning (7 citations)
- Manipulating SGD with Data Ordering Attacks (7 citations)
- Automated Discovery of Adaptive Attacks on Adversarial Defenses (4 citations)
and 8 more in the interactive map.
Black-box and transfer attacks · 11 papers
- RobustBench: a standardized adversarial robustness benchmark (114 citations)
- On Success and Simplicity: A Second Look at Transferable Targeted Attacks (45 citations)
- TRS: Transferability Reduced Ensemble via Promoting Gradient Diversity and Model Smoothness (16 citations)
- Provably Efficient Black-Box Action Poisoning Attacks Against Reinforcement Learning (12 citations)
- Adversarial Attacks on Black Box Video Classifiers: Leveraging the Power of Geometric Transformations (7 citations)
- Random Noise Defense Against Query-Based Black-Box Attacks (7 citations)
- Characterizing the risk of fairwashing (6 citations)
- Finding Optimal Tangent Points for Reducing Distortions of Hard-label Attacks (5 citations)
- Disrupting Deep Uncertainty Estimation Without Harming Accuracy (4 citations)
- A Little Robustness Goes a Long Way: Leveraging Robust Features for Targeted Transfer Attacks (2 citations)
- Meta-Learning the Search Distribution of Black-Box Random Search Based Adversarial Attacks (2 citations)
Backdoor and poisoning · 5 papers
- Backdoor Attack with Imperceptible Input and Latent Modification (36 citations)
- Anti-Backdoor Learning: Training Clean Models on Poisoned Data (35 citations)
- Adversarial Neuron Pruning Purifies Backdoored Deep Models (27 citations)
- Qu-ANTI-zation: Exploiting Quantization Artifacts for Achieving Adversarial Outcomes (3 citations)
- Excess Capacity and Backdoor Poisoning (1 citation)
