E-FreeM2: Efficient Training-Free Multi-Scale and Cross-Modal News Verification via MLLMs

2025-06-25

AI Summary

To address security risks posed by rapid misinformation propagation in mobile and wireless networks, this paper proposes a training-free, dynamically retrievable multimodal fact verification system. The method integrates pretrained vision-language models with large language models, enabling multi-scale cross-modal verification of image-text pairs through dynamic cross-modal retrieval—thereby circumventing vulnerabilities of conventional supervised models to adversarial attacks and data poisoning. Its lightweight architecture facilitates deployment on resource-constrained edge devices. Evaluated on two mainstream fact-checking benchmarks, the system achieves state-of-the-art (SOTA) performance and demonstrates significantly enhanced robustness over baseline approaches. Experimental results validate its effectiveness and security advantages in bandwidth- and compute-limited wireless environments.

Technology Category

• Computer Vision: Multi-modal Vision
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The rapid spread of misinformation in mobile and wireless networks presents critical security challenges. This study introduces a training-free, retrieval-based multimodal fact verification system that leverages pretrained vision-language models and large language models for credibility assessment. By dynamically retrieving and cross-referencing trusted data sources, our approach mitigates vulnerabilities of traditional training-based models, such as adversarial attacks and data poisoning. Additionally, its lightweight design enables seamless edge device integration without extensive on-device processing. Experiments on two fact-checking benchmarks achieve SOTA results, confirming its effectiveness in misinformation detection and its robustness against various attack vectors, highlighting its potential to enhance security in mobile and wireless communication environments.

Problem

• Detect misinformation in mobile and wireless networks
• Verify news credibility without training using multimodal models
• Enable lightweight fact-checking on edge devices

Innovation

• Training-free retrieval-based multimodal verification system
• Leverages pretrained vision-language and large language models
• Lightweight design for seamless edge device integration