Abstract: For the large redundancy and noise problem in the network data which the intrusion detection system has to deal with under the large-scale Internet environment, a hybrid intrusion detection method based on light-weighted artificial immune computation is proposed. The minimum information entropy discretization method is used to pre-process the detection data, and principal component analysis is applied to extracting the features. Negative selection algorithm is applied to online detection, and the characters are extracted for unknow or large-scale unknown connections, so the intrusion detection is realized by the artificial immune algorithm. Finally, the detector with evolutionary capacity anomaly is used to training and testing, and the extracted anomaly features are added into the fast matching database to update the database in time. Simulation results show that the algorithm can improve detection efficiency of the whole hybrid intrusion detection system and the detection velocity can satisfy the real time request.