关键词: Internet垃圾电话, 端点检测, 评分系统, 漏报率, 误报率

Abstract:

Based on the existing solution for spam over Internet telephony (SPIT) detection, a method using human conversational model to describe the call pattern of human telephone call is proposed. A conversational pattern state sequence is generated through the conversational model from a call. Via using Bayesian probability model, the possibility that the call is malicious is calculated. The possibility can be used to detect the spam phone calls generated by robot, and can be applied in our perimeter defense system. Experiments on the border defense system indicate that the false negative rate of detection is less 10% while the false positive rate remains low at about 2%. It demonstrates that conversation pattern can be used as a feature to detect behavior of SPIT as the other features.

Key words: spam over Internet telephony, endpoint detection, scoring system, false negative rate, false positive rate