Visual-Invariance-Augmented Feature Optimal Alignment for Transferable Adversarial Attacks against Closed-Source MLLMs

arXiv:2610.06977 · cs.CV, cs.AI · Submitted 2026-10-04 · Read on arXiv

cs.CV, cs.AI

Submitted: 2026-10-04

Updated: 2026-10-04

Code: https://github.com/jiaxiaojunQAQ/IAU-FOA

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Adversarial Attacks — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode

Sources

• How Robust is Google's Bard to Adversarial Image Attacks?
• A study of the effect of JPG compression on adversarial images
• Boosting Transferability in Vision-Language Attacks via Diversification along the Intersection Region of Adversarial Trajectory
• A Survey on Transferability of Adversarial Examples across Deep Neural Networks
• Countering Adversarial Images using Input Transformations
• R-Bench: Graduate-level Multi-disciplinary Benchmarks for LLM & MLLM Complex Reasoning Evaluation
• When Instructions Multiply: Measuring and Estimating LLM Capabilities of Multiple Instructions Following
• Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models
• Adversarial Attacks on Multimodal Large Language Models: A Comprehensive Survey
• Survey of Adversarial Robustness in Multimodal Large Language Models
• Nesterov Accelerated Gradient and Scale Invariance for Adversarial Attacks
• Safety of Multimodal Large Language Models on Images and Texts
• Delving into Transferable Adversarial Examples and Black-box Attacks
• DeepSeek-VL: Towards Real-World Vision-Language Understanding
• Universal Adversarial Attacks against Closed-Source MLLMs via Target-View Routed Meta Optimization
• Towards Deep Learning Models Resistant to Adversarial Attacks
• VEAttack: Downstream-agnostic Vision Encoder Attack against Large Vision Language Models
• Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
• Universal Adversarial Attack on Aligned Multimodal LLMs
• PandaGPT: One Model To Instruction-Follow Them All

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models