The Detection Model Research On Network Fault Diagnosis Based On Immune Danger

With the rapid development of the current science technology, the equipments are more complicated and have higher integration. As the platform of carrying various information, the internet has become an indispensible part in one’s life. However, when a fault happens by accident, it would influence people’s daily work, which means it is of importance to detect the reason of the faults and then process faults rapidly. The faults type is of variation and randomness, therefore, it is efficient to detect and process the network fault if we find the usual fault types by categories. The rise of fault diagnosis technology, due to the practical application requirements and the development of multidisciplinary theory, provides a new way for improving the security and dependability.The traditional immune algorithm, whose core concept is to recognize self and non-self, has its own design flaw, so it is not too qualified to recognize unknown self and non-self, which means it has lower detection rate and higher false alarm rate. Inspired by danger theory, immune system would only recognize and response the antigen in the danger area, instead of all the non-self antigens, which has higher practical operability and reduce the computational complexity of recognition and response to a great extent. While the core concept of danger theory is to recognize danger not the self and non-self, so there is no complicated mutation process in the generation of antibodies, which means the variation of self set would be a less influence for danger theory than the traditional immune algorithm. Its high fault tolerance, self-adaptive and self-regulation make the superiorit.The main research content is as follows:(1)On the basis of the further study of danger theory and the analysis of the limitations of the typical Dendritic Cell Algorithm, the concept of dynamic migration threshold is added to the algorithm, which greatly improves the recognition rate.(2)Based on the deeply research of dynamic clonal mechanism, a novel algorithm is proposed which conducts clonal variation on the mature detectors, that is to rank the detectors by affinity after each iteration, divide the mature detectors into memory unit and common unit, and then conduct different degrees of clonal variation on the mature detectors according to definite probability, which can reserves the superior individuals, optimize the worse individuals, and eventually improve the quality of mature detectors.(3)This paper combines clonal selection algorithm with the modified Dendritic Cell Algorithm, and proposes a novel network fault diagnosis model. The novel model optimizes the generation strategy of mature detectors, and the experimental result shows that better classifying effect of DCA and the excellent dynamic adaptability of dynamic clonal mechanism, the novel model improves the fault recognition rate to a great extent.