The roving proxy for SMS spam and phishing detection.

摘要

SMS messages are becoming the number one communication method in recent years. Therefore, more spammers and phishers are turning to SMS as their new attack vector. Moreover, the limited computational and battery power of mobile devices makes it di?cult to implement heavy solutions in them. Spear phishing is particularly prevalent giving these circumstances, thus, targeting a speci?c group of people using only their phone numbers, which in many cases is public records. Most of the work done towards spam and phishing messages detection, used either client side or server-side solutions. Using server-side solutions only, cannot prevent against spear phishing attacks. On the other hand, most of the work done towards client side solutions suggested the use of Naive Bayes because of its low computations, but that resulted in high false positive rates.;This study, proposes the roving proxy server to address the problem of SMS phishing in particular, and spam messages in general. First, we highlight recent mobile devices security attacks, and assess the privacy of various smartphone applications. After that, we conduct a thorough analysis of SMS spam features, and compare di?erent feature selection algorithms with di?erent classi?ers to achieve higher true positive while maintaining low false positive rates. In addition, we study the e?ect of using synthetic SMS samples to solve the problem of unbalanced datasets and to enhance the classi?cation detection rates of spam messages. Furthermore, we implement a proof of concept of the proposed roving proxy architecture, and discuss the details of di?erent layers of protection. ?nally, we present a detailed analysis of the update system parameters and requirements.