ICLR 2024 accepted papers on Adversarial robustness and security

Certified and empirical robustness · 32 papers

• Adversarial Training Should Be Cast as a Non-Zero-Sum Game (4 citations)
• DeepZero: Scaling Up Zeroth-Order Optimization for Deep Model Training (3 citations)
• A Sublinear Adversarial Training Algorithm (2 citations)
• Adversarial Training on Purification (AToP): Advancing Both Robustness and Generalization (2 citations)
• Efficient local linearity regularization to overcome catastrophic overfitting (2 citations)
• Incremental Randomized Smoothing Certification (2 citations)
• On the Over-Memorization During Natural, Robust and Catastrophic Overfitting (2 citations)
• The Effect of Intrinsic Dataset Properties on Generalization: Unraveling Learning Differences Between Natural and Medical Images (2 citations)
• A Recipe for Improved Certifiable Robustness (1 citation)
• COLEP: Certifiably Robust Learning-Reasoning Conformal Prediction via Probabilistic Circuits (1 citation)
• Conserve-Update-Revise to Cure Generalization and Robustness Trade-off in Adversarial Training (1 citation)
• DAFA: Distance-Aware Fair Adversarial Training (1 citation)

and 20 more in the interactive map.

Attacks on text and graphs · 29 papers

• Training Socially Aligned Language Models on Simulated Social Interactions (16 citations)
• An LLM can Fool Itself: A Prompt-Based Adversarial Attack (13 citations)
• Can Sensitive Information Be Deleted From LLMs? Objectives for Defending Against Extraction Attacks (4 citations)
• Robustness of AI-Image Detectors: Fundamental Limits and Practical Attacks (4 citations)
• Fusion Is Not Enough: Single Modal Attacks on Fusion Models for 3D Object Detection (3 citations)
• Chameleon: Increasing Label-Only Membership Leakage with Adaptive Poisoning (2 citations)
• Adversarial Attacks on Fairness of Graph Neural Networks (1 citation)
• LipSim: A Provably Robust Perceptual Similarity Metric (1 citation)
• LRR: Language-Driven Resamplable Continuous Representation against Adversarial Tracking Attacks (1 citation)
• Understanding Reconstruction Attacks with the Neural Tangent Kernel and Dataset Distillation (1 citation)
• Embodied Active Defense: Leveraging Recurrent Feedback to Counter Adversarial Patches
• Foundation Model-oriented Robustness: Robust Image Model Evaluation with Pretrained Models

and 17 more in the interactive map.

Backdoor and poisoning · 14 papers

• Universal Jailbreak Backdoors from Poisoned Human Feedback (8 citations)
• Efficient Backdoor Attacks for Deep Neural Networks in Real-world Scenarios (6 citations)
• Backdoor Federated Learning by Poisoning Backdoor-Critical Layers (4 citations)
• BaDExpert: Extracting Backdoor Functionality for Accurate Backdoor Input Detection (3 citations)
• Poisoned Forgery Face: Towards Backdoor Attacks on Face Forgery Detection (3 citations)
• Backdoor Secrets Unveiled: Identifying Backdoor Data with Optimized Scaled Prediction Consistency (2 citations)
• Demystifying Poisoning Backdoor Attacks from a Statistical Perspective (1 citation)
• Influencer Backdoor Attack on Semantic Segmentation (1 citation)
• Universal Backdoor Attacks (1 citation)
• Adversarial Feature Map Pruning for Backdoor
• Backdoor Contrastive Learning via Bi-level Trigger Optimization
• Rethinking Backdoor Attacks on Dataset Distillation: A Kernel Method Perspective

and 2 more in the interactive map.

Black-box and transfer attacks · 6 papers

• Rethinking Model Ensemble in Transfer-based Adversarial Attacks (6 citations)
• Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (2 citations)
• BRUSLEATTACK: A QUERY-EFFICIENT SCORE- BASED BLACK-BOX SPARSE ADVERSARIAL ATTACK
• Optimal transport based adversarial patch to leverage large scale attack transferability
• PubDef: Defending Against Transfer Attacks From Public Models
• Enhancing Transferable Adversarial Attacks on Vision Transformers through Gradient Normalization Scaling and High-Frequency Adaptation