AAAI 2023 accepted papers on Adversarial robustness and security

Attacks on text and graphs · 29 papers

• Untargeted Attack against Federated Recommendation Systems via Poisonous Item Embeddings and the Defense (66 citations)
• PLMmark: A Secure and Robust Black-Box Watermarking Framework for Pre-trained Language Models (46 citations)
• HOTCOLD Block: Fooling Thermal Infrared Detectors with a Novel Wearable Design (45 citations)
• DeAR: A Deep-Learning-Based Audio Re-recording Resilient Watermarking (41 citations)
• Pseudo Label-Guided Model Inversion Attack via Conditional Generative Adversarial Network (39 citations)
• Imperceptible Adversarial Attack via Invertible Neural Networks (34 citations)
• LeNo: Adversarial Robust Salient Object Detection Networks with Learnable Noise (31 citations)
• DisGUIDE: Disagreement-Guided Data-Free Model Extraction (21 citations)
• Attack Can Benefit: An Adversarial Approach to Recognizing Facial Expressions under Noisy Annotations (18 citations)
• Defending from Physically-Realizable Adversarial Attacks through Internal Over-Activation Analysis (15 citations)
• Purifier: Defending Data Inference Attacks via Transforming Confidence Scores (12 citations)
• Improving Adversarial Robustness with Self-Paced Hard-Class Pair Reweighting (11 citations)

and 17 more in the interactive map.

Certified and empirical robustness · 28 papers

• Delving into the Adversarial Robustness of Federated Learning (43 citations)
• Formalising the Robustness of Counterfactual Explanations for Neural Networks (20 citations)
• WAT: Improve the Worst-Class Robustness in Adversarial Training (18 citations)
• Holistic Adversarial Robustness of Deep Learning Models (17 citations)
• Adversarial Robust Deep Reinforcement Learning Requires Redefining Robustness (16 citations)
• Improving Robust Fariness via Balance Adversarial Training (11 citations)
• CSTAR: Towards Compact and Structured Deep Neural Networks with Adversarial Robustness (10 citations)
• Ensemble-in-One: Ensemble Learning within Random Gated Networks for Enhanced Adversarial Robustness (10 citations)
• Understanding and Enhancing Robustness of Concept-Based Models (10 citations)
• Feature-Space Bayesian Adversarial Learning Improved Malware Detector Robustness (9 citations)
• Neural Architecture Search for Wide Spectrum Adversarial Robustness (9 citations)
• Reliable Robustness Evaluation via Automatically Constructed Attack Ensembles (7 citations)

and 16 more in the interactive map.

Black-box and transfer attacks · 10 papers

• CodeAttack: Code-Based Adversarial Attacks for Pre-trained Programming Language Models (59 citations)
• Training Meta-Surrogate Model for Transferable Adversarial Attack (24 citations)
• SSPAttack: A Simple and Sweet Paradigm for Black-Box Hard-Label Textual Adversarial Attack (17 citations)
• Black-Box Adversarial Attack on Time Series Classification (14 citations)
• Practical Disruption of Image Translation Deepfake Networks (13 citations)
• Unfooling Perturbation-Based Post Hoc Explainers (13 citations)
• Exploring Non-target Knowledge for Improving Ensemble Universal Adversarial Attacks (12 citations)
• CDTA: A Cross-Domain Transfer-Based Attack with Contrastive Learning (11 citations)
• Defending Black-Box Skeleton-Based Human Activity Classifiers (10 citations)
• Global-Local Characteristic Excited Cross-Modal Attacks from Images to Videos (10 citations)

Backdoor and poisoning · 4 papers

• Poisoning with Cerberus: Stealthy and Colluded Backdoor Attack against Federated Learning (99 citations)
• Defending against Backdoor Attacks in Natural Language Generation (26 citations)
• Defending Backdoor Attacks on Vision Transformer via Patch Processing (17 citations)
• Poisoning-Based Backdoor Attacks in Computer Vision (8 citations)