Information society with huge number of everyday acting participants becomes valuable source of surfer behavior research. Especially web server logs can be used for discovering knowledge useful in different areas. Knowledge acquired from web server logs in order to generate solutions has to be validated. The aim of this paper is presentation of web usage mining as a rather complex process and in this context elaboration of validation model. On a basis of iterative and hybrid approach to discover user navigation patterns the concept of generated knowledge validation model is proposed. Some experiments on real website allow to define a new method of generated association rules refinement including specific knowledge validation techniques. Some constraints as discovered knowledge validation criteria are defined.