Through the in-depth study on the grid and intrusion detection technology, a credible network intrusion detection system based on grid is proposed in this paper. The system deploys in the grid environment and adopts intrusion detection technology of the improved SOM neural network. To improve data collection efficiency, a rules-based means is used in data collection and processing. For the sake of realizing load balancing among the data analyzing nodes, an integrated scheduling algorithm of the task assignment based on the resource usable degree and credibility degree is used. In order to reduce alarm quantity, an alarm synthetic algorithm based on rules of similar degree and relevance degree is adopted. Experimental results show that the intrusion detection system can not only obtain better detection efficiency but also realize the load balance of resource use.