Abstract: Mobile Ad Hoc Networks（MANETs） present more security problem than traditional network for its characteristics such as open medium,dynamic,topology distributed,cooperation and constrained capability.This paper focuses on an integrity detection model,and the supervised anomaly detection classifier of which is based on support vector machines.At the same time,three kinds of pool-based active learning algorithms applied to the model are introduced.Compared with traditional self-learning algorithm,the experiment results show that pool-based active learning algorithms can decrease the request for the number of training instances and reduce the influence on the performance of noise data.It satisfies the demand of high detection rate,high anti-noise and low overhead for MANETs intrusion detection system.