基于主题模型的猫池滥用识别算法

Discovering modem pool misuse behavior based on LDA

猫池是一种可同时支持多张手机卡的通信设备,一些组织会滥用猫池从事信用诈骗、群发垃圾短信、拨打骚扰电话等活动,使商家和个人的利益都遭受了严重损失。本文提出了一种基于主题模型的猫池识别算法,将主题模型中的词语共同出现概率问题映射为多张手机卡在时间和空间上的共同出现概率问题。实验证明该算法能够有效的发现猫池通信行为,即通信行为总是在时空上高度一致,结合猫池的其它非行为特征,可以有效发现猫池滥用行为。

Modem pool is a device equipped with multiple SIM cards slots. It can be misused for credit fraud, sending spam or making harassing phone calls. This paper proposed a method based on topic model to discover modem pool behaviors from other communications effectively. Specifically, computing the co-occurrence probability of several words is mapped into computing the co-occurrence probability of several SIM cards. The experiments showed that the method can capture the behaviors of modem pool effectively. Combining with features other than behavioral features, the algorithm can discover modem pool misuse behavior accurately.