Research On The Protocol Identification Technology Of Satellite Heterogeneous Network

In recent years, satellite network has been widely used in the fields of communication, navigation and meteorological service, due to its far communication distance and wide coverage. The increasing network applications also bring some management and security issues. On the one hand, in the application layer, the number of private protocol is growing at an explosive speed. The existing application protocol identification methods which rely on the published protocol specification are losing their efficacy gradually. The satellite network resource managements are facing new challenges. On the other hand, the investigation, monitoring and early warning satellite are playing an important role in electronic information countermeasure, the unknown of target link format brings great difficulties to further information acquisition in some non-cooperative scenarios. In view of the above two challenges, this thesis has carried out the following research based on the characteristics of the satellite network environment:In face of identification of application protocol, this thesis uses LDA(Latent Dirichlet Allocation) as feature extraction model, and then predicts protocol type of network trace. The innovation lies in: 1) Accroding to the long delay characteristics of satellite networks, this thesis proposes a decision strategy between N target protocols and non-target protocols based on the existing N classes classifier, which no need to train non-target protocals to filter non-target protocols. This strategy can improve the accuracy of the target protocols with high speed. 2) In resource optimization, considering the limitation of satellite source, this paper proposes an adaptive optimal model selection algorithm based on improved Genetic Algorithm, which can choose the optimal model parameters with fewer iterations automatically. Simulation results show that the search speed of improved Genetic Algorithm increased about 20%,reduced the system occupancy rate to some extent.In the case of reverse identification of link protocol, by comparing the applicability of Apriori algorithm and FP-Growth algorithm, this thesis uses FP-Growth (Frequent Patterns Growth) to mining frequent bit string in target link, and further identifies target link frame format, frame length field and frame sequences number field. In order to meet the need of FP-Growth, this thesis proposes a bit encoding method based on position,which can reduce the time of mining in FP-Growth compare to existing method.