An Agent-Based Message Inoculation Model For Anti-SPAM

Email has been one of the most popular communication tools for network users at present, but as follows, Spam becomes the most headache problemnull. In order to decrease the spam number in the email user's inbox, there appears lots of spam filtering technologies, such as White list and Black list technology, Rule based technology, and Content based technology. However, all these technologies train the spam filter with the original mail samples, so they can't filter the new spam. The spamers try to make new spam to enter the email user's inbox. In order to solve the problem, a new spam filtering technology has been proposed, and its' name is collaborative filtering technology. The basic idea is to union several email users, who agree to share the spam information, to deal with the new spam together.Message inoculation is one of the new collaborative spam filtering technologies. It was drafted into a message format outlined in the Internet-Draft which describes a specific MIME encoding for sending message inoculations through email. Several spam filters, such as Dspam, CRM114, SpamAssian, have brought the message inoculation. However, in those spam filters with message inoculation, the implementation involves user's complex manual operations. Users' individual preferences are also ignored. And the trust issues are not taken into consideration but the reliability of the inoculation messages themselves.Agent is a active and autonomic object, and the Muti-Agent system can solve the large-scale and complicated problems. So in this thesis, Agent technology is introduced into message inoculation. An Agent-based Message Inoculation Model (ABMIM) for anti-spam is proposed. The related issues of ABMIM are discussed, and a series of experiments are conducted.The experimental results show that the proposed model can not only accomplish message inoculation, improve the filtering accuracy, but also maintain users' preferences. After embedding CTM, the incorrect inoculation messages from malicious mail users can be reduced effectively.