Deep Learning Solutions for Phishing by URL Detection

Abstract

Abstract: In this digital age, phishing attacks are something that are quite prevalent and are on the rise. This paper explores the various avenues for detecting such kind of attacks which will pave way to mitigating such kinds of attacks in the future. We primarily focused on proving that deep learning methods are much more efficient than traditional machine learning models; for this purpose we are evaluating the performance of a traditional machine learning model namely Naive Bayes and two deep learning models which are Convolutional Neural Networks(CNN) and Recurrent Neural Networks(RNN). The process starts with normalizing the input features and then the categorical data is transformed after which the dataset containing the URLs are loaded and are preprocessed. The performance of the models was evaluated against metrics like Accuracy, Precision, Recall and F1-Score.The end results proved that CNN was able to achieve the optimal performance and was capable of outperforming the other two models. Therefore this paper is of the view that such CNN or Neural Network empowered Models are the only way to mitigate these types of attacks and will also act as a catalyst in developing systems or models that are immune to such kinds of attacks.