Debate-Driven Multi-Agent LLMs for Phishing Email Detection
Open accessAbstract
Phishing attacks remain a critical cybersecurity threat. Attackers constantly refine their methods, making phishing emails harder to detect. Traditional detection methods, including rule-based systems and supervised machine learning models, either rely on predefined patterns like blacklists, which can be bypassed with slight modifications, or require large datasets for training and still can generate false positives and false negatives. In this work, we propose a multi-agent large language model (LLM) prompting technique that simulates debates among agents to detect whether the content presented on an email is phishing. Our approach uses two LLM agents to present arguments for or against the classification task, with a judge agent adjudicating the final verdict based on the quality of reasoning provided. This debate mechanism enables the models to critically analyze contextual cue and deceptive patterns in text, which leads to improved classification accuracy. The proposed framework is evaluated on multiple phishing email datasets and demonstrate that mixed -agent configurations consistently outperform homogeneous configurations. Results also show that the debate structure itself is sufficient to yield accurate decisions without extra prompting strategies.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Debate among phishing-classification agents reduces false negatives.
“debate among phishingclassification agents reduces false negatives [90]”
From Agentic AI and Large Language Models for Autonomous IoT Cybersecurity: A Systematic Survey, Taxonomy, and Research Roadmap (2026) · cites this paper as [90]
Cites
Still being checked.
Cited by
Sorted by citations
PhishDebate: An LLM-Based Multi-Agent Framework for Phishing Website Detection
Li, W., Manickam, S., Chong, Y.-W., & Karuppayah, S. (2025) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata66926.2025.11401440 · Cited by 12 · see citing works
Kytidou, E., Tsikriki, T., Drosatos, G., & Rantos, K. (2025) · Intelligent Decision Technologies · doi:10.1177/18724981251366763 · Cited by 8 · see citing works
PRIME: A Phishing Detection Framework With Quantitative and Fuzzy-Based Dual Validation
Tian, Y., Qu, Y., Ding, M., Liu, S., Tsai, P.-W., & Zhang, J. (2025) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2025.3609320 · Cited by 2 · see citing works
Nageshwaran, V., & Ezekiel, S. (2026) · Electronics · doi:10.3390/electronics15122740 · Cited by 4 · see citing works
Deep Learning-Based Intrusion Detection Systems for Phishing Email Detection: A Short Survey
Nardin, A. D., Zottin, S., Piciarelli, C., & Foresti, G. L. (2025) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw69036.2025.00784 · Cited by 1 · see citing works
Multi-Agent Debate Strategies: Survey, Taxonomy, and Challenges
Motger, Q., Oriol, M., Marco, J., & Franch, X. (2026) · arXiv · doi:10.48550/arxiv.2607.26212
Prompt Injection Attacks on LLM-Based Spam Filters
Borkowski, G. (2025) · Wroclaw University of Economics and Business eBooks · doi:10.15611/2025.07.4.02
Lee, K.-H., Lee, Y., Jeong, J., Choi, Y.-H., & Shin, D. (2026) · Electronics · doi:10.3390/electronics15102196
Abir, C. M., Das, P., Jayaraman, V., Parthasarathy, S., & Biswas, S. (2026) · International Conference on Secure Information Systems and Technologies (ICSIST) · doi:10.1109/icsist70215.2026.11706387
Nandhini, S., D, S. M., & Jegathambal, P. M. G. (2026) · International Journal of Engineering and Management Research · doi:10.31033/ijemr/16.5.2026.1951
- Authors
- Vy Nguyen, N. T., Childress, F. D., & Yin, Y.
- Affiliations
- Earlham College,Department of Computer Science,Richmond,IN,USA
- Earlham College (US)
- Venue
- 2025 13th International Symposium on Digital Forensics and Security (ISDFS)
- Publisher
- IEEE
- Published
- 2025-04-24
- Pages
- 1-5
- OpenAlex
- W4411205956
- Cited by
- 8
- References
- 0
- Citations by year
- 2026: 3 · 2025: 5
- FWCI
- 9.58
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- International Symposium on Digital Forensics and Security (ISDFS) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Information Systems › Spam and Phishing Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
- Keywords
- phishing email detection, multi-agent large language models, LLM prompting, adversarial reasoning, email classification
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
