A Dynamically Selected GPT Model for Phishing Detection

This paper introduces a novel approach to augmenting Incident Response Teams (IRT) by leveraging fine-tuned GPT models for early threat detection. Traditional IRTs often face challenges in timely response, prompting the need for automation. Our solution focuses on automating the pre-detection phase...

Full description

Saved in:
Bibliographic Details
Main Authors: Beydemir, Alp Baris, Sezgin, Ulas, Dogan, Umutcan, Asiklar, Burak Engin, Yerlikaya, Fahri Anil, Bahtiyar, Serif
Format: Conference Proceeding
Language:English
Subjects:
Automation
Computational modeling
Cybersecurity
Electronic mail
Faces
Generative AI
Information technology
Large language models
Natural Language Processing
Phishing
Phishing Detection
Threat assessment
Time factors
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper introduces a novel approach to augmenting Incident Response Teams (IRT) by leveraging fine-tuned GPT models for early threat detection. Traditional IRTs often face challenges in timely response, prompting the need for automation. Our solution focuses on automating the pre-detection phase by alerting users about potentially harmful emails before they are opened, addressing the issue of insufficient response time. In comparison to the base model, our fine-tuned GPT models exhibit superior performance. The results of this study will be forwarded to the IRT for further evaluation and potential integration into a pre-detection system. Notably, our method emphasizes content and context analysis of emails, crucial for identifying insider threats. By employing Generative Large Language Models (GLLM), specifically tuned for this purpose, we aim to enhance the detection capabilities, contributing to a more robust incident response strategy in cybersecurity.
ISSN:2770-5226
DOI:10.1109/ACIT62333.2024.10712553