Artificial immune system (AIS) is considered as an adaptive computational intelligence method that could be used for detecting and preventing current computer network threats. AIS generates Antibodies (self) competent in recognizing Antigen (non-self), which is considered as an anomaly technique. This paper aims to develop artificial immune system (AIS) that consists of two levels. Level one is developed using Genetic Algorithm, while level two is developed using C4.5 decision tree algorithm. The proposed system trained with clustered features that are selected from NSL-KDD cup data-set. Each level produces two antibodies (that could recognize Normal and Antigen access-records). The recognition accuracy of the developed system reaches 96%. The behavior of each level is studied. The best feature-set that suits each level is specified.