This paper puts forward a designing method of Dependable Network System Security Risk Diagnosis (DSSRD). It gets reduced information table, which implies that the number of evaluation criteria is reduced with no information loss, and then, this table is used to develop classification rules and infer appropriate parameters. It’s capable of overcoming several shortcomings in existing diagnosis methods, such as a dilemma between stability and redundancy. The average speed of training in DSSRD is almost twice fast as that in SMO. The experiment implemented by this method shows a good diagnostic ability.