Pith. sign in

REVIEW 3 cited by

Seeing is Deceiving: Exploitation of Visual Pathways in Multi-Modal Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2411.05056 v1 pith:MBN5BMEU submitted 2024-11-07 cs.CR cs.AI

Seeing is Deceiving: Exploitation of Visual Pathways in Multi-Modal Language Models

classification cs.CR cs.AI
keywords visualmllmsmulti-modaldatalikemodelsriskssecurity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Multi-Modal Language Models (MLLMs) have transformed artificial intelligence by combining visual and text data, making applications like image captioning, visual question answering, and multi-modal content creation possible. This ability to understand and work with complex information has made MLLMs useful in areas such as healthcare, autonomous systems, and digital content. However, integrating multiple types of data also creates security risks. Attackers can manipulate either the visual or text inputs, or both, to make the model produce unintended or even harmful responses. This paper reviews how visual inputs in MLLMs can be exploited by various attack strategies. We break down these attacks into categories: simple visual tweaks and cross-modal manipulations, as well as advanced strategies like VLATTACK, HADES, and Collaborative Multimodal Adversarial Attack (Co-Attack). These attacks can mislead even the most robust models while looking nearly identical to the original visuals, making them hard to detect. We also discuss the broader security risks, including threats to privacy and safety in important applications. To counter these risks, we review current defense methods like the SmoothVLM framework, pixel-wise randomization, and MirrorCheck, looking at their strengths and limitations. We also discuss new methods to make MLLMs more secure, including adaptive defenses, better evaluation tools, and security approaches that protect both visual and text data. By bringing together recent developments and identifying key areas for improvement, this review aims to support the creation of more secure and reliable multi-modal AI systems for real-world use.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. GUIGuard-Bench: Toward a General Evaluation for Privacy-Preserving GUI Agents

    cs.CR 2026-01 unverdicted novelty 8.0

    GUIGuard-Bench is a new benchmark with annotated GUI screenshots that measures privacy recognition, planning fidelity under protection, and utility impact for trajectory-based GUI agents.

  2. ORCA: An Agentic Reasoning Framework for Hallucination and Adversarial Robustness in Vision-Language Models

    cs.CV 2025-09 unverdicted novelty 6.0

    ORCA is an inference-time agentic framework that boosts LVLM accuracy on hallucination benchmarks by 3.64-40.67% and adds adversarial robustness via cross-model validation with small vision tools.

  3. ORCA: An Agentic Reasoning Framework for Hallucination and Adversarial Robustness in Vision-Language Models

    cs.CV 2025-09 unverdicted novelty 6.0

    ORCA is an agentic reasoning framework that enhances factual accuracy and adversarial robustness of pretrained LVLMs via an Observe-Reason-Critique-Act loop with small vision models, reporting accuracy gains of up to ...