AAAI 2022 accepted papers on Adversarial robustness and security
47 of the 1,319 papers accepted at AAAI 2022 (3.6%) are on Adversarial robustness and security, compared with 3.4% at AAAI 2021. They fall into 4 topics.
See AAAI 2022 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 · 19 papers
- CMUA-Watermark: A Cross-Model Universal Adversarial Watermark for Combating Deepfakes (118 citations)
- FCA: Learning a 3D Full-Coverage Vehicle Camouflage for Multi-View Physical Adversarial Attack (115 citations)
- Hard to Forget: Poisoning Attacks on Certified Machine Unlearning (64 citations)
- Robust Heterogeneous Graph Neural Networks against Adversarial Attacks (64 citations)
- TextHoaxer: Budgeted Hard-Label Adversarial Attacks on Text (49 citations)
- DeepAuth: A DNN Authentication Framework by Model-Unique and Fragile Signature Embedding (31 citations)
- Learning Universal Adversarial Perturbation by Adversarial Example (30 citations)
- Adversarial Training for Improving Model Robustness? Look at Both Prediction and Interpretation (25 citations)
- Synthetic Disinformation Attacks on Automated Fact Verification Systems (24 citations)
- Attacking Video Recognition Models with Bullet-Screen Comments (22 citations)
- CLPA: Clean-Label Poisoning Availability Attacks Using Generative Adversarial Nets (21 citations)
- The King Is Naked: On the Notion of Robustness for Natural Language Processing (15 citations)
and 7 more in the interactive map.
Certified and empirical robustness · 17 papers
- Improved Text Classification via Contrastive Adversarial Training (95 citations)
- Consistency Regularization for Adversarial Robustness (51 citations)
- PatchUp: A Feature-Space Block-Level Regularization Technique for Convolutional Neural Networks (35 citations)
- Efficient Robust Training via Backward Smoothing (22 citations)
- Adversarial Robustness in Multi-Task Learning: Promises and Illusions (20 citations)
- CC-CERT: A Probabilistic Approach to Certify General Robustness of Neural Networks (19 citations)
- Combating Adversaries with Anti-adversaries (19 citations)
- Improved Gradient-Based Adversarial Attacks for Quantized Networks (16 citations)
- Input-Specific Robustness Certification for Randomized Smoothing (16 citations)
- Robust Optimal Classification Trees against Adversarial Examples (16 citations)
- Adversarial Bone Length Attack on Action Recognition (15 citations)
- Improving Bayesian Neural Networks by Adversarial Sampling (15 citations)
and 5 more in the interactive map.
Black-box and transfer attacks · 8 papers
- Towards Transferable Adversarial Attacks on Vision Transformers (111 citations)
- Sparse-RS: A Versatile Framework for Query-Efficient Sparse Black-Box Adversarial Attacks (103 citations)
- Boosting the Transferability of Video Adversarial Examples via Temporal Translation (43 citations)
- Making Adversarial Examples More Transferable and Indistinguishable (29 citations)
- Learning to Learn Transferable Attack (17 citations)
- Adversarial Learning from Crowds (16 citations)
- Blindfolded Attackers Still Threatening: Strict Black-Box Adversarial Attacks on Graphs (13 citations)
- Natural Black-Box Adversarial Examples against Deep Reinforcement Learning (11 citations)
