Research On Situation - Aware Technology Of Information Content Security Event In Telecommunication Network

With the growing of telecommunication network scale and the rapid development of communication technology, telecommunication network emerges an unprecedented flourishing trend in the number of users and penetration、types of terminal、types of service and the degree of integration with Internet. However, it also brings new challenges to the telecommunication network information content security. The Information Content Security Incidents(ICSI) such as harassing audio/video calls and spam messages/multimedia messages gradually increased, have been influenced and interfered with the users’ work and life, and bring instability factors to the development of harmonious society. Therefore, the control and management of such incidents have become a research hotspot in telecommunications network information content security. Information Content Security Incidents Situation Awareness(ICSISA) describes the element of perception, comprehension and prediction which cause the change of the incidents situation. It can provide effective and convincing data to the decision making, and it has an important practical significance and theoretical value.The research object of this paper is the ICSI in telecommunication network, including the harassing audio/video calls and spam messages/multimedia messages; Taking the situation awareness of ICSI in telecommunication as the goal of research, which represented by harassing audio/video calls and spam messages/multimedia messages; Following the classic framework of situation awareness proposed by Endsly, and dividing the information processing of ICSISA into situation perception, situation comprehension and situation prediction, it is the main thread of this article. Among them, the situation comprehension includes event classification and situation evaluation.The main contents and innovations are as follows:1、A situation perception method based on association rules is proposed. Using the difference of communication characteristics between ICSI and conventional communication to mine the relative data from behavior、 relation、 location and content characteristics to realize the discovery of ICSI; Aiming at the false alarm problem in big data, an inspection standard is proposed based on Bonferroni correction. It can filer the frequent items whether they meet the conditions of the occurrence or not. Experimental results demonstrate that the proposed method is effective, and it has a great detection rate in the case of low false alarm rate and missing alarm rate;2、Distributed Power Set Apriori(DPS-Apriori) algorithm and Hierarchical Co-operative Co-evolution Genetic(HCCG) algorithm are put forward to resolve the problem of associate mining in offline and online data. DPS-Apriori algorithm uses power method to generate sub item sets of frequent one item sets which filtered by one support level, and takes them as candidate items. This method reduces the number of scanning datasets times and cutting times. It can improve the calculation speed and make the results more complete by maintaining the frequent item sets which may be filtered by the origin algorithm. Experimental results demonstrate that the algorithm’s running time is short and has strong ability in parallel computing. It has better performance than Apriori algorithm; The HCCG algorithm has hierarchical structure, uses genetic algorithm for each sub-species populations formed by item sets to evolution as a local solution, uses the co-evolution to achieve information’s interaction and transmission in dominate populations, which will integrate local solutions to obtain global solution. Experimental results demonstrate that under the premise of guaranteed accuracy, the algorithm runs faster, and has strong focusing ability、good generalization and strong ability in parallel computing than the existed algorithm;3 、 An event classification based on Multi-dimension Information Association LDA(MIA-LDA) model is proposed. It uses the communication time character as timeline, and uses LDA model to calculate the results of classification of user’s information and communication content characters in each time slice. After the similarity metric calculating of the classification results, merges them with the results of update data. It can get the result of current on-line events classification. Experimental results demonstrate that the model has good ability in generalization and classification. It can realize the on-line classificy of information content security incidents. Compareing with the other algorithms, the proposed algorithm can classify the incidents in shorter time and has higher accuracy;4、A hierarchical situation evaluation model is proposed. The model uses hierarchical structure and calculates the situation assessment values of incident level、area level and system level separately. Among them, incidents level is based on behavioral and content characteristics in incidents characteristics; area level is according to the relationship and location characteristics; system level integrates the involved area levels. The method of the calculation of parameters is defined. Experimental results demonstrate that the model and method is feasible and reliable. It can reflect the impaction of incidents effectively and grasp the changing of the incidents;5、A situation prediction method based on Elitist Model Immune Genetic Algorithm(E-IGA) is proposed to optimize the structure and parameters of RBF neural network. The model uses elite selection strategy and ensures good genes to be retained into the next generation. At the same time, it increases the diversity of variation to a certain extent through the disturbance of the annealing factor. And the model improves the convergence rate and local search capacity of the whole algorithm. Experimental results demonstrate that the algorithm can predict the situation of ICSI accurately, and has high degree of fitting with the actual data. It reflects the trend of the incidents intensity effectively. The algorithm has strong convergence and lesser training time. Its overall performance has advantages when compared with existed algorithms.