Over the years series of attacks has been launched by attackers in order to steal from users, one of the most prevalent attacks is the phishing attacks, this phishing attacks is has become alarming, various approaches has been used to curb this menace but non has actually provide the best solution to this threatening cybercrime such as phishing and therefore the need for better system tools to counter such approaches is paramount. This research proposes the use of deep learning techniques using Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and hybrid models of CNN-GRU in detecting phishing domains. The efficient feature extraction abilities of CNNs help in the usage of URL structures where spatial relations exist, whilst the sequential structural formulation of domain names is modeled using GRUs. CNN-GRU model Their integration allows the local pattern detection of the CNN to be meshed with the long sequence utilization of GRU for better accuracy. All models are subjected to standard datasets for phishing with conclusive performance metrics such as Detection Rate (DR), Precision, Recall, F1-Score, and Accuracy. The system utilized small and large datasets. In this regard, a hybrid CNN-GRU model did show better resilience against phishing attempts than its singular counterparts on a small dataset with an accuracy of 0.9440 while CNN 0.9337 and GRU 0.9402. For the larger datasets CNN outperformed the other two models with an accuracy of 0.9382 while GRU had 0.9026 and CNN-GRU settled for 0.9261. The primary aim of this paper is to evaluate standalone models and hybridized model on two datasets (small and large) to know which of datasets they can perform better in detecting phishing attacks.

Authors: T. J. Ayo, O. D. Alowolodu

Published in: International Conference for Internet Technology and Secured Transactions (ICITST-2024)

• Date of Conference: 4-6 November 2024
• DOI: 10.20533/ICITST.2024.0018
• ISBN: 978-1-913572-76-1
• Conference Location: St Anne’s College, Oxford University, UK