Attack model BBFPAN based on fuzzy Petri net has some drawback such as the lack of learning mechanism.A delaminating algorithm which can partition the BBFPAN into several levels is presented.This algorithm can be applied to the study about the BBFPAN using the neural network.In order to make fine description to the network attacks and finding the weakness of the network system,the basic inference rules about BBFPAN is put forward which originally applies Both-Branch Fuzzy Logic into the attack modeling.After this problem,based on the basic inference rules,a delaminating reasoning algorithm is proposed which can enrich the reasoning method of the attack model BBFPAN and the effectiveness of the algorithm is verified by experiment.