Research on Android Malicious URL Detection Based on Machine Learning

Abstract

AbstractIn recent years, malicious URL attacks against mobile users are increasing, the detection of malicious URLs is of great significance to defend against network attacks. The traditional research on mobile malicious URL detection has the defects of high false positive rate and false positive rate. In response to these questions, this paper mainly constructs a cross-platform applications to classify malicious URLs based on blacklist technology and support vector machines. Moreover, a malicious URL detection system is designed and developed on Android to achieve accurate detection of malicious domain names such as Trojans and viruses. Finally, lots of experiments are done by testing the data of 9174 domain names. The accuracy rate is 98.25%, which improves the security of mobile terminals and web application as a whole.KeywordsURL detectionSVM algorithmMachine learningFeature extractionModel training