Saved in:
Bibliographic Details
Main Authors: Dalmiere, Antony, Auriol, Guillaume, Nicomette, Vincent, Marchand, Pascal
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.22515
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913916713435136
author Dalmiere, Antony
Auriol, Guillaume
Nicomette, Vincent
Marchand, Pascal
author_facet Dalmiere, Antony
Auriol, Guillaume
Nicomette, Vincent
Marchand, Pascal
contents Traditional phishing detection often overlooks psychological manipulation. This study investigates using Large Language Model (LLM) In-Context Learning (ICL) for fine-grained classification of phishing emails based on a taxonomy of 40 manipulation techniques. Using few-shot examples with GPT-4o-mini on real-world French phishing emails (SignalSpam), we evaluated performance against a human-annotated test set (100 emails). The approach effectively identifies prevalent techniques (e.g., Baiting, Curiosity Appeal, Request For Minor Favor) with a promising accuracy of 0.76. This work demonstrates ICL's potential for nuanced phishing analysis and provides insights into attacker strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2506_22515
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle In-context learning for the classification of manipulation techniques in phishing emails
Dalmiere, Antony
Auriol, Guillaume
Nicomette, Vincent
Marchand, Pascal
Cryptography and Security
Artificial Intelligence
Traditional phishing detection often overlooks psychological manipulation. This study investigates using Large Language Model (LLM) In-Context Learning (ICL) for fine-grained classification of phishing emails based on a taxonomy of 40 manipulation techniques. Using few-shot examples with GPT-4o-mini on real-world French phishing emails (SignalSpam), we evaluated performance against a human-annotated test set (100 emails). The approach effectively identifies prevalent techniques (e.g., Baiting, Curiosity Appeal, Request For Minor Favor) with a promising accuracy of 0.76. This work demonstrates ICL's potential for nuanced phishing analysis and provides insights into attacker strategies.
title In-context learning for the classification of manipulation techniques in phishing emails
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2506.22515