Detecting attacks and Fraud links in Google Play

Abstract

Malicious URLs are wide accustomed mount numerous cyber attacks together with spamming, phishing and malware. Detection of malicious URLs and identification of threat varieties area unit important to thwart these attacks. Knowing the kind of a threat allows estimation of severity of the attack and helps adopt an efficient countermeasure. Existing strategies generally sight malicious URLs of one attack kind. during this paper, we tend to propose technique using machine learning to sight malicious URLs of all the popular attack varieties and establish the character of attack a malicious URL makes an attempt to launch. Our method uses a spread of discriminative options together with textual properties, link structures, webpage contents, DNS information, and network traffic. several of those features area unit novel and extremely effective. Our experimental studies with forty,000 benign URLs and thirty two,000 malicious URLs obtained from real-life web sources show that our technique delivers a superior performance: the accuracy was over ninety eight in police work malicious URLs and over 93% in characteristic attack varieties. we tend to conjointly report our studies on the effectiveness of every cluster of discriminative features, and discuss their evadability.