IWSPA-2023-Adversarial-Synthetic-Dataset

This repository contains two phishing/legitimate dataset extensions from IWSPA 2.0 dataset:

1. Adversarial: This dataset is consisting of successful attack examples for both phishing and legitimate emails, generated by four different types of text attack techniques (Textfooler, PWWS, DeepWordBug, and BAE) from the TextAttack framework on the fine-tuned ALBERT phishing detection model.
2. Synthetic: This dataset is consisting of synthetic phishing email examples using a fine-tuned GPT-2 model.

Citing Dataset

If you use IWSPA-2023-Adversarial-Synthetic-Dataset for your research, please cite [Adversarial Robustness of Phishing Email Detection Models] (https://dl.acm.org/doi/abs/10.1145/3579987.3586567) paper

@inproceedings{mehdi2023adversarial, title={Adversarial Robustness of Phishing Email Detection Models}, author={Mehdi Gholampour, Parisa and Verma, Rakesh M}, booktitle={Proceedings of the 9th ACM International Workshop on Security and Privacy Analytics}, pages={67--76}, year={2023} }