A method of collaborative learning intrusion detection applied in data grid, which draws advantages of current popular technologies like distributed detection in intrusion detection domain and ensemble learning etc., combines method of abnormal detection and feature detection, adopts BP neural network collaborative learning for complementary advantages, and makes intrusion detection system based on the method applies to data grid environment. Deploy a plurality of strong detectors integrated by BP neural network in center-nod according to security requirement of different kinds of nods in data grid by the invention, and collect new intrusion sample feature come from everywhere whenever possible to make sure about the security of important copy in center-nod. Deploy collaborative intrusionanalysis engine in center-nod meanwhile to provide collaborative detection service for all normal nods. A plurality of normal nods are organized by center-nod for collaborative calculation and suspected data can be detected which cannot be judged by original signal nod so as to improve security of normal nod.