| Summary: | One of the most challenging categories of threats circulating in the online world is social engineering, with malicious web links, fake news, clickbait, and other tactics. Malware URLs are extremely dangerous because they represent the main propagating vector for web malware. Malicious web links detection is a challenging task because the detection mechanism should not influence the consumers’ online experience. The proposed solutions must be sensitive enough, and fast enough to perform the detection mechanism before the user accesses the link and downloads its content. Our paper proposes three goals. The main purpose of this paper is to refine a methodology for malicious web links detection that may be used to experiment with machine learning algorithms. Moreover, we propose to use this methodology for training and comparing several machine learning algorithms such as Random Forest, Decision Tree, K-Nearest Neighbor. The results are compared, justified, and placed in the malicious web links literature. In addition, we propose to identify the most relevant features and draw some observations about them.
|
|---|