A collaborative and multi-agent approach to e-mail filtering

CAFE (collaborative agents for filtering e-mails) is a multi-agent system to collaboratively filter spam from users´ mail. CAFE associates each user with a proxy agent that acts as an interface between the user´s e-mail client and the server. With the support of other agents, the proxy agent classifies messages into three categories: ham (good messages), spam and spam-presumed. The system analyzes each message using three approaches: the first is based on an hash function, the second (static) on DNSBL (DNS-based black lists) databases and the third (dynamic) on a Bayesian filter. We give a mathematical representation of the system, showing that if the users collaborate, the fault probability decreases in proportion to the number of active users.