Abstract: In network environment, junk email always acts as a threat to the security of the mail system and takes up a lot of network resources. To deal with this problem, considering the properties of spam body, the probability distribution of spam keywords was studied based on Bayesian theory. A spam filtering algorithm was put forward by means of SMTP protocol on the Solaris platform and LAN emulation test was performed. The results show that the spam hit rate is increased with the application of Bayesian spam probability formula.