ICML 2024 accepted papers on Adversarial robustness and security
64 of the 2,610 papers accepted at ICML 2024 (2.5%) are on Adversarial robustness and security, compared with 3.0% at ICML 2023. They fall into 4 topics.
See ICML 2024 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Certified and empirical robustness · 30 papers
- Robust Classification via a Single Diffusion Model (4 citations)
- OODRobustBench: a Benchmark and Large-Scale Analysis of Adversarial Robustness under Distribution Shift (3 citations)
- Adversarial Robustness Limits via Scaling-Law and Human-Alignment Studies (2 citations)
- Robust Universal Adversarial Perturbations (2 citations)
- Enhancing Adversarial Robustness in SNNs with Sparse Gradients (1 citation)
- Et Tu Certifications: Robustness Certificates Yield Better Adversarial Examples (1 citation)
- Towards Optimal Adversarial Robust Q-learning with Bellman Infinity-error (1 citation)
- Attack-free Evaluating and Enhancing Adversarial Robustness on Categorical Data
- Breaking the Barrier: Enhanced Utility and Robustness in Smoothed DRL Agents
- Can Implicit Bias Imply Adversarial Robustness?
- Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval
- Compositional Curvature Bounds for Deep Neural Networks
and 18 more in the interactive map.
Attacks on text and graphs · 20 papers
- HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal (23 citations)
- Image Hijacks: Adversarial Images can Control Generative Models at Runtime (4 citations)
- Be Your Own Neighborhood: Detecting Adversarial Examples by the Neighborhood Relations Built on Self-Supervised Learning (2 citations)
- Membership Inference Attacks on Diffusion Models via Quantile Regression (2 citations)
- Purify Unlearnable Examples via Rate-Constrained Variational Autoencoders (2 citations)
- Robust Stable Spiking Neural Networks (2 citations)
- DataFreeShield: Defending Adversarial Attacks without Training Data (1 citation)
- IOI: Invisible One-Iteration Adversarial Attack on No-Reference Image- and Video-Quality Metrics (1 citation)
- RAUCA: A Novel Physical Adversarial Attack on Vehicle Detectors via Robust and Accurate Camouflage Generation (1 citation)
- Rethinking Adversarial Robustness in the Context of the Right to be Forgotten (1 citation)
- Revisiting Character-level Adversarial Attacks for Language Models (1 citation)
- Data Poisoning Attacks against Conformal Prediction
and 8 more in the interactive map.
Backdoor and poisoning · 8 papers
- IBD-PSC: Input-level Backdoor Detection via Parameter-oriented Scaling Consistency (3 citations)
- Generalization Bound and New Algorithm for Clean-Label Backdoor Attack (1 citation)
- TERD: A Unified Framework for Safeguarding Diffusion Models Against Backdoors (1 citation)
- Better Safe than Sorry: Pre-training CLIP against Targeted Data Poisoning and Backdoor Attacks
- A Theoretical Analysis of Backdoor Poisoning Attacks in Convolutional Neural Networks
- Causality Based Front-door Defense Against Backdoor Attack on Language Models
- Defense against Backdoor Attack on Pre-trained Language Models via Head Pruning and Attention Normalization
- Energy-based Backdoor Defense without Task-Specific Samples and Model Retraining
Black-box and transfer attacks · 6 papers
- CosPGD: an efficient white-box adversarial attack for pixel-wise prediction tasks (13 citations)
- Stealing part of a production language model (7 citations)
- BadPart: Unified Black-box Adversarial Patch Attacks against Pixel-wise Regression Tasks (1 citation)
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (1 citation)
- Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution (1 citation)
- Towards Efficient Training and Evaluation of Robust Models against $l_0$ Bounded Adversarial Perturbations