Malicious Website Detection Through Deep Learning Algorithms

Traditional methods that detect malicious websites, such as blacklists, do not update frequently, and they cannot detect new attackers. A system capable of detecting malicious activity using Deep Learning (DL) has been proposed to address this need. Starting from a dataset that contains both malevolent and benign websites, classification is done by extracting, parsing, analysing, and preprocessing the data. Additionally, the study proposes a Feed-Forward Neural Network (FFNN) to classify each sample. We evaluate different combinations of neurons in the model and perform in-depth research of the best performing network. The results show up to 99.88% of detection of malicious websites and 2.61% of false hits in the testing phase (i.e. malicious websites classified as benign), and 1.026% in the validation phase.