Abstract: This paper presents a novel IDS mechanism based on the theory of data fusion-DFIDS, which is designed, according to intrusion patterns with complex features, to heighten the accuracy of the system while detecting complex intrusion acts. The performance of the whole detecting system is enhanced because DFYDS effectively overcomes the limitations of conventional single sensor detecting system regarding aspects of data collecting and analyzing by using optimized model of the fusion of parallel-distributed detection and decision. In this paper, a theoretical model of DFIDS is established and compared with traditional IDS, which proves that DFIDS is of better performance as to detecting accuracy.