Salvato in:
Dettagli Bibliografici
Autori principali: Su, Guangzhi, Huang, Shuchang, Ke, Yutong, Liu, Zhuohang, Qian, Long, Huang, Kaizhu
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2510.26830
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912678482542592
author Su, Guangzhi
Huang, Shuchang
Ke, Yutong
Liu, Zhuohang
Qian, Long
Huang, Kaizhu
author_facet Su, Guangzhi
Huang, Shuchang
Ke, Yutong
Liu, Zhuohang
Qian, Long
Huang, Kaizhu
contents Multimodal large language models (MLLMs) have achieved impressive performance across diverse tasks by jointly reasoning over textual and visual inputs. Despite their success, these models remain highly vulnerable to adversarial manipulations, raising concerns about their safety and reliability in deployment. In this work, we first generalize an approach for generating adversarial images within the HuggingFace ecosystem and then introduce SmoothGuard, a lightweight and model-agnostic defense framework that enhances the robustness of MLLMs through randomized noise injection and clustering-based prediction aggregation. Our method perturbs continuous modalities (e.g., images and audio) with Gaussian noise, generates multiple candidate outputs, and applies embedding-based clustering to filter out adversarially influenced predictions. The final answer is selected from the majority cluster, ensuring stable responses even under malicious perturbations. Extensive experiments on POPE, LLaVA-Bench (In-the-Wild), and MM-SafetyBench demonstrate that SmoothGuard improves resilience to adversarial attacks while maintaining competitive utility. Ablation studies further identify an optimal noise range (0.1-0.2) that balances robustness and utility.
format Preprint
id arxiv_https___arxiv_org_abs_2510_26830
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SmoothGuard: Defending Multimodal Large Language Models with Noise Perturbation and Clustering Aggregation
Su, Guangzhi
Huang, Shuchang
Ke, Yutong
Liu, Zhuohang
Qian, Long
Huang, Kaizhu
Machine Learning
Cryptography and Security
Multimodal large language models (MLLMs) have achieved impressive performance across diverse tasks by jointly reasoning over textual and visual inputs. Despite their success, these models remain highly vulnerable to adversarial manipulations, raising concerns about their safety and reliability in deployment. In this work, we first generalize an approach for generating adversarial images within the HuggingFace ecosystem and then introduce SmoothGuard, a lightweight and model-agnostic defense framework that enhances the robustness of MLLMs through randomized noise injection and clustering-based prediction aggregation. Our method perturbs continuous modalities (e.g., images and audio) with Gaussian noise, generates multiple candidate outputs, and applies embedding-based clustering to filter out adversarially influenced predictions. The final answer is selected from the majority cluster, ensuring stable responses even under malicious perturbations. Extensive experiments on POPE, LLaVA-Bench (In-the-Wild), and MM-SafetyBench demonstrate that SmoothGuard improves resilience to adversarial attacks while maintaining competitive utility. Ablation studies further identify an optimal noise range (0.1-0.2) that balances robustness and utility.
title SmoothGuard: Defending Multimodal Large Language Models with Noise Perturbation and Clustering Aggregation
topic Machine Learning
Cryptography and Security
url https://arxiv.org/abs/2510.26830