With recent advances in network based technology needs protecting computers and networks which becomes a huge problem. Based on information coming from various response teams a computer was attacked or broken into more than once per second. In this paper, two grains levels intrusion detection system (IDS) is suggested fine-grained and coarse-grained. In normal case the intrusions are not detected, to improve the performance the most suitable IDS level is the coarse-grained. Any intrusion is detected by coarse-grained IDS after that the fine-grained is used to detect the possible attack details. Very fast decision tree (VFDT) algorithm is used in both of these detection levels. In order to ensure efficiency of the proposed model, it has been tested on KDD CUP 99 dataset and a real traffic dataset. Experimental results demonstrate that the proposed model is highly successful in detecting known and unknown attacks
