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

TL;DR

E-FreeM2 is a training-free, multimodal fact verification system that uses pretrained models and dynamic data retrieval to effectively detect misinformation on mobile devices, offering robustness and efficiency.

Contribution

It introduces a novel training-free, retrieval-based multimodal verification approach leveraging pretrained models for secure, lightweight misinformation detection.

Findings

Achieves state-of-the-art results on two fact-checking benchmarks.

Demonstrates robustness against adversarial attacks and data poisoning.

Enables seamless edge device deployment without extensive training.

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.