"VNeSafe: Machine Learning-Assisted System for Detecting Malicious URLs and Spam Calls"

Abstract

Spam calls and malicious Uniform Resource Locators (URLs) have become major concerns for Internet users. Phishing, spam, and drive-by-download attacks can be initiated by malicious URLs, while normal users may experience irritation from spam calls. To tackle the aforementioned issues, we provide VNeSafe, a machine learning-assisted system, in this paper. By leveraging user feedback, VNeSafe may identify a phone number that is spam. Particularly, it keeps track of how many times a phone subscriber has been reported as spam. When such a number is over a predetermined threshold, VNeSafe automatically adds the phone number to a blacklist and blocks it. Furthermore, VNeSafe uses a natural language processing technique named TF-IDF in order to extract good features from a URL. The Random Forest algorithm then makes use of these features to determine whether the URL is malicious or not. Our empirical research has demonstrated that Random Forest can offer a real-time detection with an F1-score of 0.9298. This algorithm is ready to be deployed in VNeSafe and used on a general mobile device.