This paper presents a system for automatically detecting and filtering unsolicited electronic messages. The underlying filtering method is based on email origin and content. A heuristic knowledge base formed by spam words is extracted from labelled emails by a finite state automata. The processing of three parts of every email by a single Bayesian filter and the integration of the every part classification allows to achieve a maximum performance goal. The system is dynamic and interactive and evolves from the evolution of spam by incremental machine learning.