论文部分内容阅读
为了解决网络电话(VoIP)业务面临的Internet垃圾电话(SPIT)问题,在针对SPIT检测方案的基础上提出利用人类会话模型对会话进行建模,在该模型的基础上将会话转化为表征会话模式的状态序列,再利用朴素贝叶斯分类器计算得到该会话模式为恶意会话模式可能性的数值度量.该数值度量可用作检测呼叫机器人制造垃圾电话的依据,并在VoIP边界防护系统中得到了应用.实验结果表明,该方法在较低误报率(2%左右)的情况下能保证较低的漏报率(小于10%),同时证明了会话模式可作为判断SPIT的特性之一.
In order to solve the problem of Internet Spam (SPIT) faced by the VoIP business, this paper proposes a human session model based on the SPIT detection scheme to model the session. Based on this model, the session is transformed into a token session mode , And then use the Naive Bayesian classifier to calculate the value of the probability that the conversation pattern is a malicious conversation pattern.The numerical metric can be used as a basis for detecting the call robot’s manufacturing junk phone and obtained in the VoIP border protection system The experimental results show that this method can guarantee a lower false negative rate (less than 10%) under the condition of lower false alarm rate (about 2%), and at the same time it is proved that the session mode can be used as one of the characteristics of judging SPIT .