An Experimental Analysis of Transformer-Based Models for Phishing Detection

DOI

10.2991/978-94-6463-950-6_8How to use a DOI?

Keywords

Email; Phishing; Transformer

Abstract

Phishing remains one of the most prominent cybersecurity threats, and email classification systems become critical for online protection. This paper proposes a transformer-based approach for phishing email detection that attains high classification accuracy with real-time efficiency and deployability. By utilizing deep learning methods and HuggingFace Transformers library, several models such as DistilBERT, RoBERTa, ELECTRA, MobileBERT, ConvBERT, and ALBERT are compared on a benchmark phishing email dataset. DistilBERT shows the best inference time (0.0042s). ELECTRA yields the best accuracy (98.20%) and F1 (0.98), although with the largest model size (1254.07 MB) and is best applied in accuracy-sensitive use cases. MobileBERT yields the smallest model size (93.91 MB) and is best applied in environments with limited resources. ConvBERT and ALBERT models are a well-balanced trade-off between accuracy and efficiency. The double-metric assessment ensures not only the accuracy of the models but also their feasibility for deployment. This transformer-based phishing detection system provides a scalable, reliable, and flexible method to improve email security while satisfying multiple performance and infrastructure requirements.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Jyoti Madake
AU  - Sanika Kadam
AU  - Sayali Kadam
AU  - Anushka Khandelwal
PY  - 2025
DA  - 2025/12/29
TI  - An Experimental Analysis of Transformer-Based Models for Phishing Detection
BT  - Proceedings of the International Conference on Smart Systems and Social Management (ICSSSM 2025)
PB  - Atlantis Press
SP  - 92
EP  - 103
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6463-950-6_8
DO  - 10.2991/978-94-6463-950-6_8
ID  - Madake2025
ER  -