CVPR 2025 accepted papers on Adversarial robustness and security
63 of the 2,871 papers accepted at CVPR 2025 (2.2%) are on Adversarial robustness and security, compared with 2.3% at CVPR 2024. They fall into 4 topics.
See CVPR 2025 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
Attacks on text and graphs · 34 papers
- Anyattack: Towards Large-scale Self-supervised Adversarial Attacks on Vision-language Models (16 citations)
- Instant Adversarial Purification with Adversarial Consistency Distillation (12 citations)
- Adv-CPG: A Customized Portrait Generation Framework with Facial Adversarial Attacks (9 citations)
- On the Zero-shot Adversarial Robustness of Vision-Language Models: A Truly Zero-shot and Training-free Approach (8 citations)
- PatchGuard: Adversarially Robust Anomaly Detection and Localization through Vision Transformers and Pseudo Anomalies (6 citations)
- R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning (6 citations)
- Steering Away from Harm: An Adaptive Approach to Defending Vision Language Model Against Jailbreaks (6 citations)
- STEREO: A Two-Stage Framework for Adversarially Robust Concept Erasing from Text-to-Image Diffusion Models (6 citations)
- Detecting Adversarial Data Using Perturbation Forgery (5 citations)
- Divide and Conquer: Heterogeneous Noise Integration for Diffusion-based Adversarial Purification (5 citations)
- Seeing is Not Believing: Adversarial Natural Object Optimization for Hard-Label 3D Scene Attacks (5 citations)
- TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models (5 citations)
and 22 more in the interactive map.
Black-box and transfer attacks · 13 papers
- Chain of Attack: On the Robustness of Vision-Language Models Against Transfer-Based Adversarial Attacks (12 citations)
- Improving Transferable Targeted Attacks with Feature Tuning Mixup (10 citations)
- Improving Adversarial Transferability on Vision Transformers via Forward Propagation Refinement (9 citations)
- Improving the Transferability of Adversarial Attacks on Face Recognition with Diverse Parameters Augmentation (9 citations)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (8 citations)
- Mind the Gap: Detecting Black-box Adversarial Attacks in the Making through Query Update Analysis (5 citations)
- From Head to Tail: Efficient Black-box Model Inversion Attack via Long-tailed Learning (4 citations)
- NoPain: No-box Point Cloud Attack via Optimal Transport Singular Boundary (4 citations)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (3 citations)
- Data-free Universal Adversarial Perturbation with Pseudo-semantic Prior (2 citations)
- Derivative-Free Diffusion Manifold-Constrained Gradient for Unified XAI
- Jailbreaking the Non-Transferable Barrier via Test-Time Data Disguising
and 1 more in the interactive map.
Certified and empirical robustness · 9 papers
- CLIP is Strong Enough to Fight Back: Test-time Counterattacks towards Zero-shot Adversarial Robustness of CLIP (8 citations)
- TAET: Two-Stage Adversarial Equalization Training on Long-Tailed Distributions (7 citations)
- Certified Human Trajectory Prediction (4 citations)
- MOS-Attack: A Scalable Multi-objective Adversarial Attack Framework (3 citations)
- Gain from Neighbors: Boosting Model Robustness in the Wild via Adversarial Perturbations Toward Neighboring Classes (2 citations)
- Two is Better than One: Efficient Ensemble Defense for Robust and Compact Models (1 citation)
- Rethinking the Adversarial Robustness of Multi-Exit Neural Networks in an Attack-Defense Game
- Stop Walking in Circles! Bailing Out Early in Projected Gradient Descent
- Doppelgangers and Adversarial Vulnerability
Backdoor and poisoning · 7 papers
- Infighting in the Dark: Multi-Label Backdoor Attack in Federated Learning (8 citations)
- Revisiting Backdoor Attacks against Large Vision-Language Models from Domain Shift (8 citations)
- BadToken: Token-level Backdoor Attacks to Multi-modal Large Language Models (7 citations)
- Invisible Backdoor Attack against Self-supervised Learning (7 citations)
- UIBDiffusion: Universal Imperceptible Backdoor Attack for Diffusion Models (4 citations)
- Where the Devil Hides: Deepfake Detectors Can No Longer Be Trusted (3 citations)
- DeDe: Detecting Backdoor Samples for SSL Encoders via Decoders (2 citations)