Search

Software-Defined Networking (SDN) has emerged as a novel network architecture for facilitating and simplifying network control and management. The main fundamental of SDN is the separation of the control and data planes that allows to rapidly, simply manage and configure network operations. However, because of the logically centralized control plane, SDN brings many security challenges, especially to be the victim of Controller-aimed Distributed Denial of Service (DDoS) attacks. In this paper, we proposed a solution to detect and mitigate this dangerous threat to protect SDN controllers. Our proposal consists of two components including a network application for supporting in decision making and a network function for enforcing the detection and mitigation tasks. We also contribute a multi-layer attack detection mechanism and a three-phase mitigation approach to treat with the attacks. Our methodology is implemented on OpenDayLight controller and evaluated using a high-speed test-bed network. The results show that our solution is able to detect attacks after 40 milliseconds on average, and the accuracy of detection process is around 95%. Moreover, it also can effectively, efficiently mitigate attacks to reduce CPU Utilization from high (approx. 90%) to remain average (approx. 20%).

PPC-Tree and N-list have been proven to be very efficient and been used in mining frequent itemsets widely. The main problem of the novel structures is that the way of First Constructing Last Encoding method is adopted in the tree-building phase. This causes excessive time consumption to mine frequent itemsets. In this paper, we propose SFO-Set based on SFO-Tree, a more efficient data structure, to mine frequent itemsets. SFO-Tree construction employs the Constructing and Encoding method in comparison with PPC-Tree. Based on SFO-Sets and Subsume Index, FI_SS algorithm, an efficient hybrid method for mining frequent itemsets, is proposed. For evaluating the performance of FI_SS, we conduct lots of experiments on a variety of public datasets. Experimental results show that FI_SS algorithm has advantages in running time.

This paper analyzes the bottleneck of EJB system performance, proposes an idea of optimizing EJB design pattern to improve the performance, and then give a layered and integrated optimization method based on mixed design patterns. Accordingly, the performance of the application was improved and the system has better flexibility and maintainability. Experiments show the improvement of performance at last.