Caller-Centrality: Identifying Telemarketers in a VoIP Network

Abstract

In recent years, VoIP (Voice over Internet Protocol) has emerged as cheap telephony medium for a long distance international and domestic calls. The number of unwanted calls from telemarketers and scammers has also risen recently, because of VoIP telephony that makes easier to initiate large number of calls without being tracing back by authorities. It is utmost important for the VoIP operators to gain trust of their customers by blocking telemarketers and scammers at the edge of the network. To address this challenge, in this paper, we present a system called Caller-Centrality that effectively identifies and blocks telemarketers/spammers without being intrusive to the caller and the callee. Caller-Centrality first models the user relationships as a caller graph and then computes reputation of the caller using weighted centrality measure. The edge weights between caller and the callee are assigned from call rate and call duration between caller and the callee. We evaluated our approach anonymized real-data set collected from a small VoIP operator. The evaluation results reveal that Caller-Centrality successfully identifies suspected telemarketers.