There is no standard solution we can use to completely protect against computer network intrusion. Every solution has its advantages and drawbacks. Soft computing is considered as a promising paradigm to cope with the dynamic evolution of networks. In previous works, we presented two soft computing approaches of intrusion detection. The first one is based on the neuro-fuzzy and the second one is based on the genetic fuzzy one. In this work, we elaborate an empirical comparative study to highlight the benefits of each method in intrusion detection and exploit their complementarities to enhance the detection rate of all types of attacks as well as decrease the false positives rate.