These days, internet has become rich of information. This information should be properly managed. People are using email services but most of the emails are irrelevant or junk called spams. So a proper filter for these emails should be applied on email servers. Bayesian approach is being popular for filtering spams. This approach is based on the bayes method. Currently available spam filters are static in nature. Results of spam filtering can be unexpectedly better than earlier if a dynamic nature is used with available spam filters. In this paper Bayesian spam filter is discussed and some techniques for improvement of Bayesian filter are proposed.