Abstract
We address the challenge of detecting exaggeration in cybersecurity tweets on X, where misinformation spreads rapidly. Our novel framework uses local Large Language Models (LLMs) and Retrieval Augmented Generation (RAG) to gather evidence and assess tweets' rhetorical intensity, offering graded exaggeration scores. Validated through a human study and a pilot that matches LLM results with human labels, this work lays the groundwork for improved misinformation detection tools.
| Original language | English |
|---|---|
| Title of host publication | 3D-Sec 2025 - Proceedings of the 1st ACM Workshop on Deepfake, Deception and Disinformation Security |
| Editors | Simon S. Woo, Shahroz Tariq, Sharif Abuadbba, Kristen Moore, Tim Walita, Mario Fritz, Bimal Viswanath |
| Publisher | Association for Computing Machinery, Inc |
| Pages | 1-4 |
| Number of pages | 4 |
| ISBN (Electronic) | 9798400719028 |
| DOIs | |
| State | Published - 12 Oct 2025 |
| Event | 1st ACM Workshop on Deepfake, Deception and Disinformation Security, 3D-Sec 2025 - Taipei, Taiwan, Province of China Duration: 13 Oct 2025 → 17 Oct 2025 |
Publication series
| Name | 3D-Sec 2025 - Proceedings of the 1st ACM Workshop on Deepfake, Deception and Disinformation Security |
|---|
Conference
| Conference | 1st ACM Workshop on Deepfake, Deception and Disinformation Security, 3D-Sec 2025 |
|---|---|
| Country/Territory | Taiwan, Province of China |
| City | Taipei |
| Period | 13/10/25 → 17/10/25 |
Bibliographical note
Publisher Copyright:© 2025 Copyright held by the owner/author(s).
Keywords
- Cybersecurity
- Fake news
- LLM
- Misinformation detection
ASJC Scopus subject areas
- Computer Science Applications
- Computer Vision and Pattern Recognition
- Computer Graphics and Computer-Aided Design
- Computer Networks and Communications