Chinese Spam Filter under Adversarial Impact
Yaqing Zhao, Yan Lei Xu, Xiaodan Zhao · 2015
Machining learning techniques have achieved great success in anti-spam area.But because of the limitations of these techniques, classifiers derived from them often get attacked by spam senders thus posing a threat to the whole Spam filtering system.This article briefly describes the type of attacks to classifiers and then simulates an attack on a public Chinese spam corpus to analyze the adversarial impact of several major classifiers.