A Transformer Model Approach for Robust Phishing Detection
Open Access DepositedThis research uses a quantitative cross-sectional study to examine conventional models and traditional machine learning algorithm models in detecting email phishing. The escalating threat of phishing, exemplified by the staggering volume of 3.4 billion daily phishing emails in 2022 (Baki & Verma, 2022), poses a significant risk to individuals and businesses, resulting in an alarming $8.8 billion in losses according to the Federal Trade Commission. In response, various deep learning models have emerged to enhance threat detection. However, a critical research gap exists in exploring the benefits of generative pre-training transformer models in email phishing detection. Catching these emails is crucial, but existing methods have limitations. Traditional models are effective on their own but using Natural Language Processing (NLP) methods such as transformers (BERT, GPT-2) in ensemble models may be even better. Characterizing and investigating the performance of email phishing detection ensemble models with transformers is an essential exploration when combating email phishing attacks. In response to the negative impacts of email phishing, deep learning-based approaches developed to improve phishing email detection (Mughaid et al., 2022). Among, deep learning-based approaches, LSTM and GRU offer “long-term dependencies” and the “ability to handle sequential data” (Do et al., 2022; Halgaš et al., 2020; Sun et al., 2021). Multimodel approaches have also been introduced as a means of mitigating limitations identified in different deep-learning models (e.g. Roy et al., 2022). However, research on ensembled model approach using generative pre-training transformers-based models is needed to understand the characteristics of each approach and their associated outcomes. Research limitations do not explore the possibility of ensemble transformer models.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.