Optimization Research In A Muti-Objective Model For Integrating Network Security And QoS

In network application systems, network security and quality of service(QoS) consume system resource and may significantly impact system performance. In order to ensure better system performance under resource-constrained system(environment), it is important to make a balance between network security and QoS performance.This author had researched the history, development and achievements on optimization of network security and QoS performance in network service. Due to the limitation of previous research and the demand of differentiated services, a multi-objective optimization model-- Muti-Objective Optimal(MOO) model is proposed. In this model, which is based on a platform of database systems, intrusion detection and prevention systems(IDPSs) are set up to protect database system security. Using the quantitative evaluation, a set of IDPSs configuration is obtained to help balancing network security and QoS of system. By introducing the expected utility theory, the security and QoS quantitative functions in MOO model is revised, which makes the model more in line with actual reqirements of the users. Taking into account the factors of service charging in the real-life environment, the traditional security and QoS optimization problem is extended to increase both the accounting optimization feature and proposed accounting quantitative function. It helps this MOO model providing value-added service. According to different combinations of IDPSs configuration, these three quantitative functions can calculate each optimization objective value for the current configuration, which quantitatively reflects the interaction relationship among these optimization objectives. In order to obtain a set of optimal solutions, multi-objective evolutionary algorithms are used as optimation search algorithm in this MOO model. In optimization applications with MOO model, NSGA-II algorithm is used to solve the two objectives- security and QoS optimization problem. And MOEA/D algorithm is used to solve the three objectives- security, QoS and accounting optimization problem. During the evaluation stage, with the quantitative functions and multi-objective evolutionary algorithm, MOO model searches for available combinations of IDPSs configuration to obtain a set of near-best solutions. Then, depending on the requirement of user’s preference, one of near-best solution is selected as the actual IDPSs configuration.Meanwhile, considering the elitism in the original NSGA-II is not suitable for two objectives optimization problem in the MOO model, the framework of NSGA-II is revised and a binary crossover approach for the MOO model is proposed. The new NSGA-II is similar in algorithm complexity as original NSGA-II. The simulation shows that, the optimal results would more closer to the best result when using the new NSGA-II. In other words, the new NSGA-II could found better optimal solutions at lower security demands than the original NSGA-II.