ICCV 2025 accepted papers on Adversarial robustness and security
65 of the 2,701 papers accepted at ICCV 2025 (2.4%) are on Adversarial robustness and security, compared with 2.4% at ICCV 2023. They fall into 4 topics.
See ICCV 2025 in the interactive map
Submitting to ICCV? Checklist →
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 · 31 papers
- Exploring the Adversarial Vulnerabilities of Vision-Language-Action Models in Robotics (55 citations)
- Jailbreaking Multimodal Large Language Models via Shuffle Inconsistency (50 citations)
- Heuristic-Induced Multimodal Risk Distribution Jailbreak Attack for Multimodal Large Language Models (43 citations)
- One Perturbation is Enough: On Generating Universal Adversarial Perturbations against Vision-Language Pre-training Models (43 citations)
- IDEATOR: Jailbreaking and Benchmarking Large Vision-Language Models Using Themselves (36 citations)
- FaceShield: Defending Facial Image against Deepfake Threats (9 citations)
- ODDR: Outlier Detection & Dimension Reduction Based Defense Against Adversarial Patches (8 citations)
- Web Artifact Attacks Disrupt Vision Language Models (8 citations)
- Membership Inference Attacks with False Discovery Rate Control (5 citations)
- Revisiting Adversarial Patch Defenses on Object Detectors: Unified Evaluation, Large-Scale Dataset, and New Insights (5 citations)
- AutoPrompt: Automated Red-Teaming of Text-to-Image Models via LLM-Driven Adversarial Prompts (4 citations)
- GLEAM: Enhanced Transferable Adversarial Attacks for Vision-Language Pre-training Models via Global-Local Transformations (4 citations)
and 19 more in the interactive map.
Certified and empirical robustness · 12 papers
- Adversarial Robust Memory-Based Continual Learner (7 citations)
- SAFER: Sharpness Aware layer-selective Finetuning for Enhanced Robustness in vision transformers (5 citations)
- Adversarial Training for Probabilistic Robustness (4 citations)
- Failure Cases Are Better Learned But Boundary Says Sorry: Facilitating Smooth Perception Change for Accuracy-Robustness Trade-Off in Adversarial Training (3 citations)
- CIARD: Cyclic Iterative Adversarial Robustness Distillation (2 citations)
- Confound from All Sides, Distill with Resilience: Multi-Objective Adversarial Paths to Zero-Shot Robustness (2 citations)
- Mitigating Catastrophic Overfitting in Fast Adversarial Training via Label Information Elimination (2 citations)
- Towards Adversarial Robustness via Debiased High-Confidence Logit Alignment (2 citations)
- Adversarial Robustness of Discriminative Self-Supervised Learning in Vision (1 citation)
- FastJSMA: Accelerating Jacobian-based Saliency Map Attacks through Gradient Decoupling
- NAPPure: Adversarial Purification for Robust Image Classification under Non-Additive Perturbations
- Scaling and Taming Adversarial Training with Synthetic Data
Backdoor and poisoning · 11 papers
- Backdoor Mitigation by Distance-Driven Detoxification (4 citations)
- Stealthy Backdoor Attack in Federated Learning via Adaptive Layer-wise Gradient Alignment (4 citations)
- Backdooring Self-Supervised Contrastive Learning by Noisy Alignment (3 citations)
- CopyrightShield: Enhancing Diffusion Model Security Against Copyright Infringement Attacks (2 citations)
- Prototype Guided Backdoor Defense via Activation Space Manipulation (2 citations)
- Backdoor Attacks on Neural Networks via One-Bit Flip (1 citation)
- SPD: Shallow Backdoor Protecting Deep Backdoor Against Backdoor Detection (1 citation)
- Backdoor Defense via Enhanced Splitting and Trap Isolation
- DISTIL: Data-Free Inversion of Suspicious Trojan Inputs via Latent Diffusion
- Seal Your Backdoor with Variational Defense
- StolenLoRA: Exploring LoRA Extraction Attacks via Synthetic Data
Black-box and transfer attacks · 11 papers
- Boosting Adversarial Transferability via Residual Perturbation Attack (5 citations)
- Boosting Generative Adversarial Transferability with Self-supervised Vision Transformer Features (4 citations)
- Leveraging Spatial Invariance to Boost Adversarial Transferability (3 citations)
- PLA: Prompt Learning Attack against Text-to-Image Generative Models (3 citations)
- SMP-Attack: Boosting the Transferability of Feature Importance-based Adversarial Attack with Semantics-aware Multi-granularity Patchout (3 citations)
- Enhancing Adversarial Transferability by Balancing Exploration and Exploitation with Gradient-Guided Sampling (2 citations)
- Enhancing Transferability of Targeted Adversarial Examples via Inverse Target Gradient Competition and Spatial Distance Stretching (2 citations)
- ViT-EnsembleAttack: Augmenting Ensemble Models for Stronger Adversarial Transferability in Vision Transformers (2 citations)
- SAM Encoder Breach by Adversarial Simplicial Complex Triggers Downstream Model Failures (1 citation)
- Boosting Adversarial Transferability via Negative Hessian Trace Regularization
- KOEnsAttack: Towards Efficient Data-Free Black-Box Adversarial Attacks via Knowledge-Orthogonalized Substitute Ensembles
