Premium
Detection and prevention of spam over Internet telephony in Voice over Internet Protocol networks using Markov chain with incremental SVM
Author(s) -
Vennila G.,
Manikandan M. S. K.,
Suresh M. N.
Publication year - 2016
Publication title -
international journal of communication systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.344
H-Index - 49
eISSN - 1099-1131
pISSN - 1074-5351
DOI - 10.1002/dac.3255
Subject(s) - computer science , the internet , markov chain , classifier (uml) , support vector machine , voice over ip , overhead (engineering) , computer network , markov model , reputation , machine learning , artificial intelligence , world wide web , social science , sociology , operating system
Summary This paper is mainly focused on the detection and prevention of the spammers such as telemarketing callers in Voice over Internet Protocol networks. The existing spam over Internet telephony (SPIT) detection mechanisms use call characteristics features such as reputation rate, call rejection rate, and user feedback that might increase the computation overhead and communication overhead. In this paper, a 2‐tier model is proposed and implemented for detecting, preventing, and mitigating SPIT callers. The 2‐tier model consists of a Markov chain (MC) and an incremental support vector machine (ISVM). The MC, in the first tier, detects the telemarketing callers, whereas the ISVM classifier in the second tier segregates these callers from the legitimate users. Also, the ISVM gradually mitigates the arrival of telemarketing callers at the recipient side. The performance of the proposed MC‐ISVM classifier is tested using an experimental test bed, and the results show that the 2‐tier model reports a promising blockage probability rate of 0.9 against spammers with a false positive rate of less than 0.01.