A Convolutional Neural Network Model to Detect Illegitimate URLs

摘要

The internet has become a global means of communication in all aspects of our lives. Using the internet for social activities, financial transactions, e-banking and shopping have become simple and safe to a high extent. Global online retail business is growing dramatically every year. On the other side, illegitimate fraud websites have been growing exponentially due to the increased number of e-businesses and internet users, hence, the plethora of fraud detection systems. Fraudsters always try to trap internet users to collect and steal their sensitive information using tricky techniques. Phishing is one way to trick internet users and direct them to access false Uniform Resource Locators (URLs) in an attempt to steal usernames, passwords and credit cards details. Discovering an intelligent way to determine false websites from true ones is a challenging problem. In this paper, we propose a deep learning convolutional neural network (CNN-1D) model to detect illegitimate URLs. To evaluate the performance of the model, we carried out few experiments using a benchmarked dataset. We used two evaluation measures: accuracy and the area under the receiver operating characteristic (ROC) curve (AUC). The proposed CNN-1D model was able to achieve good performance for predicting the unseen URLs and detecting illegitimate websites. In the testing phase, the classifier achieved an accuracy rate of 94.31% and an overall performance (AUC) rate of 91.23%.