Statistical Rules for Thai Spam Detection

In this paper, we propose statistical rules for Thai spam detection. Our approach is to generate Thai rules for SpamAssassin which is a popular spam detection system. We combine the advantage of rule-based and statistical-based methods. Rules can be shared, but it must be updated frequently to cope with spammers' tactics. Statistical filter can adapt to new types of spam with few human intervention by retraining the misclassify messages. The knowledge of statistical filter is usually large and limited to a server. However, our Thai rules, inducted from statistical method, can easily be shared and can cope with new variations of spam messages. The results show that Thai rules can filter spam more efficiently.