AAAI 2021 accepted papers on Adversarial robustness and security
56 of the 1,654 papers accepted at AAAI 2021 (3.4%) are on Adversarial robustness and security, compared with 3.0% at AAAI 2020. They fall into 4 topics.
See AAAI 2021 in the interactive map
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
Attacks on text and graphs · 24 papers
- Amnesiac Machine Learning (202 citations)
- Adversarial Training with Fast Gradient Projection Method against Synonym Substitution Based Text Attacks (65 citations)
- Initiative Defense against Facial Manipulation (64 citations)
- Improving Robustness to Model Inversion Attacks via Mutual Information Regularization (63 citations)
- Enabling Fast and Universal Audio Adversarial Attack Using Generative Model (56 citations)
- Curse or Redemption? How Data Heterogeneity Affects the Robustness of Federated Learning (51 citations)
- Exacerbating Algorithmic Bias through Fairness Attacks (40 citations)
- How Does Data Augmentation Affect Privacy in Machine Learning? (40 citations)
- Composite Adversarial Attacks (36 citations)
- Universal Adversarial Perturbations Through the Lens of Deep Steganography: Towards a Fourier Perspective (36 citations)
- Token-Aware Virtual Adversarial Training in Natural Language Understanding (34 citations)
- UAG: Uncertainty-aware Attention Graph Neural Network for Defending Adversarial Attacks (25 citations)
and 12 more in the interactive map.
Certified and empirical robustness · 24 papers
- Intrinsic Certified Robustness of Bagging against Data Poisoning Attacks (82 citations)
- Understanding Catastrophic Overfitting in Single-step Adversarial Training (75 citations)
- Detecting Adversarial Examples from Sensitivity Inconsistency of Spatial-Transform Domain (61 citations)
- Adversarial Robustness through Disentangled Representations (24 citations)
- Attribute-Guided Adversarial Training for Robustness to Natural Perturbations (24 citations)
- Stable Adversarial Learning under Distributional Shifts (22 citations)
- Exploring the Vulnerability of Deep Neural Networks: A Study of Parameter Corruption (18 citations)
- Simple and Effective Stochastic Neural Networks (18 citations)
- Amata: An Annealing Mechanism for Adversarial Training Acceleration (17 citations)
- Efficient Certification of Spatial Robustness (15 citations)
- FIMAP: Feature Importance by Minimal Adversarial Perturbation (15 citations)
- Beating Attackers At Their Own Games: Adversarial Example Detection Using Adversarial Gradient Directions (12 citations)
and 12 more in the interactive map.
Black-box and transfer attacks · 5 papers
- Generating Natural Language Attacks in a Hard Label Black Box Setting (96 citations)
- Adversarial Defence by Diversified Simultaneous Training of Deep Ensembles (19 citations)
- PID-Based Approach to Adversarial Attacks (15 citations)
- Improving Adversarial Robustness via Probabilistically Compact Loss with Logit Constraints (12 citations)
- Error-Correcting Output Codes with Ensemble Diversity for Robust Learning in Neural Networks (10 citations)
