Abstract: How to effectively generate detectors is one of the important problems of negative selection algorithm and its practicability. In this paper, a new algorithm which adaptively generates detectors is proposed. This algorithm can regulate current detector set according to actual circumstances, and quickly detect abnormal changes in a very large non-self space by using a small detector set. This algorithm has a good generality to some degree in a number of anomaly detection applications. Its theoretical foundation is analyzed. An experimental example and results are given to prove that this algorithm is effective.