SOCIO-LENS: Spotting Unsolicited Caller Through Network Analysis

Spam and unwanted content has been a significant challenge for the Internet technologies (email, social networks, search engines, etc.) for decades. However, in recent years, the advent of modern and cheap telephony technologies and larger user base (more than six billion users) has attracted scammers to use telephony for distributing unwanted content via instant messaging and calls. Detection of unwanted caller in the telephony has become challenging because the content is available only after the call has already been answered by the recipients and thus is too late to block the unwanted caller after the call has already been established. One of the interesting possibilities is to develop a telephony blacklist database using social behaviour of users towards their friends and family circle by modelling call meta-data as a weighted network graph. In this chapter, we model user’s behaviour as a weighted call graph network and identify malicious users by analysing different network features of users. To this extent, we have identified a set of features that help represent malicious and non-malicious behaviour of users in a network. We have conducted rigorous experimentation of the proposed system via its implementation with data set collected by small-scale telecommunication operator. We present the outcomes of our evaluation highlighting the efficacy of the system’s performance and identifying possible directions for future work.