Dynamic Policy Access Model Based On Concept Lattice

With the development and maturity of wireless technology, supportive to wireless has inevitably become necessary for next generation network. And mobile IP network will surely surpass the telecom as well as other protocol network to be the mainstream of future network. However, the practical application and popularization of mobile IP network will not come into being until the QoS of wireless access network can be guaranteed. As to the QoS of wireless access network, how to ensure the QoS during handoff of mobile host is another sticking point. Nowadays, policy based network management has grown up to a brand new network management mode, if applied to the management of QoS and resource in mobile IP network, it will sure help a lot with relative work. Still, policy based dynamic resource allocation during the process of mobile IP handoff needs to create new policy according to network condition dynamically as well as pertinent parameters. Here comes the requirement: an appropriate policy access model has to be built, which not only possesses the function of the policy storage but also can provide great convenience for policy conflict detection and resolution, for the later are two key steps on the way of putting policy network management into realization.According to the policy management system, policies are memorized in policy depository which can either be relational database or directory server, even files with certain format like XML file is permitted for policy storage. However, some born shortcomings of document like forbidden of writing and reading at the same time, slow access and restrict of document size, these flaws enough stop document from playing a role in policy access. Although with large numbers of products like SQL Server, DB2, Informix and Oracle, relational database technology can still find its embarrass when applied to policy access. Different from relational database, directory server does not support transaction disposal needed by batch updating, and only can execute simple updating action, however, it is more appropriate than relational database when comes to large amount of data searching, not to mention its ability of information copy function, which not only decrease the system response time but increases the usability and reliability of the system. Nevertheless, when faced with dynamic network whose conditions change with unstable environment, both directory server and relational database fail to satisfy the requirements of the changing network, for the future network, as long as real management based on policy, dynamically created policies are of necessary, which sure brings forward higher demands for the policy access system. Since not only searching but conflict detection and resolution happen to the system, especially the conflict detection part, with key step of finding whether conditions of different policies intersect, but present policy access systems do not even mention this important function. While, concept lattice does a good job here, its lattice structure reveals the relationship of generalization and specification, and it has a great advantage in judging whether conditions of different policies intersect.Policy rule consists of condition and action, and the condition part is an expression of attributes. For certain attribute, its values possess partial orderings. So, if proper formal context is defined and concept lattice is created through formal concept analysis, the partial ordering relationship between attribute values is achieved, use the concept lattice as the policy depository, and organize policies into lattice, first, concept lattice supplies with us more semantic information, its context relation skillfully deals with intersection judgment, on the other hand, because of the stability of background knowledge of policies that we use to build the concept lattice, this system changes little and certainly, we can say that this system also bears stability. So to sum up, through this concept lattice model, policies are classified effectively; also a good foundation is laid for later policy conflict detection.Based on concept lattice based dynamic policy access model, a conflict detection pre-process algorithm is proposed, according to concept lattice, super-classification concept and sub-classification concept intersects in their extensions, which means attributes values overlap, so with rich built-in semantics of policy rule attributes values overlap, through taking advantage of this, judging intersection between attribute values can be transferred to searching super-concept and sub-concept. the conflict detection pre-process algorithm based on concept lattice provides with conflict detection semantic structure, these greatly reduce the number of policies necessary for conflict detection and increase conflict detection efficiency, it has prepared a good foundation for future policy conflict detection.Further more, consider optimization of the above algorithm, if the whole work is done when the system is running, then it will sure take too much time and hinder the efficiency of the system, so if possible, some work can be done off-line, this certainly takes some burden off from on-line processing. The realization of the algorithm itself needs to be improved also, from the relative chapter we know, the main part of demanded results have been gained at the off-line process, so when needed, just a traversal of the hash table is enough to get conflict policy set. From experiments, these improvements upgrade the applicability of the system.The experimental results show that the new model is effective, and the algorithm based on it is of great importance in reducing the number of policy rules to be detected during conflict detection.