Abstract: We analyze the characteristic of intrusion attack based on access control granularity and mult-i dimension secure topological space,and propose a spatial expended-dimension character based intrusion detection model－SEDIDS.We construct the semantic net model for the object of access control system,and detect the integrity of access control object by the complete reasoning on semantic net,which is used as the evidence of intrusion detection.The proposed model replaces the traditional method which only depends on the comparison of system access profile constructed by training dataset.The experimental results show that this model obtains low false positive rate and low false negative rate,and is of high efficiency.