In view of video frequency frame data in time axis which has Gaussian distribution characteristic, a clustering algorithm based on the measure of Gauss kernel function similarity and an Improved Particle Swarm Optimization(IPSO) is presented. The proposed algorithm can realize accuracy clustering by Gauss kernel function similarity measure, and speed up the clustering process by the IPSO. Experimental results show that the proposed algorithm has greater searching capability and clustering accuracy, which is superior to the C-Mean in analysis of video frequency frame data clustering.
Based on the principle of Computational Auditory Scene Analysis(CASA), this paper describes a model of speech separation by analyzing onset and offset of auditory events. The model detects onsets and offsets, and generates segments by matching corresponding onset and offset fronts. The model is evaluated with three kinds of corpus, evaluation shows that it can separate all the corpus with excellent performance.
For the load balancing problem of cluster system, an improved dynamic load balancing strategy based on feedback is introduced. Combined with periodic feedback theory, the strategy divides update interval into several equal subintervals, and adopts a random mechanism to distribute requests in each subinterval. The model of the process of cluster system is founded based on queuing theory and the formula for calculating the system performance is obtained. Simulation results show that, compared with Aggressive LI strategy, this strategy can attain equivalent performance at the expense of smaller cost, and has small complexity of implementation.
In order to solve the problem of grid resource discovery, a grid resource discovery method is presented which is adjusted by preference partition and M-Flooding algorithm. The method gives a new method of measuring resource similarity degree and an improved message diffusion way M-Flooding algorithm. The grid spatial nodes are divided into different groups according to their respective preference attributes. Searching requests are transmitted in groups, which avoids the disadvantages brought by the traditional blind search method. Experimental result shows the method improves the efficiency of grid resource discovery and reduces the average path length of resource discovery.
This paper analyzes some problems of existing network storage system and presents a basic architecture for out-of-band virtual network storage system based on Enterprise Volume Management System(EVMS) technology. The client, the internal read and write operation, the online expansion of storage capacity, the load balancing, and the asynchronous metadata update algorithms for this out-of-band virtual storage system are designed and implemented.
This paper presents a network traffic matrix estimation method. In the situation of giving the network topology and link traffic, it computes the information entropy according to the link traffic, and utilizes Expectation Maximization(EM) algorithm to compute the traffic demand of the pair of Origin and Destination(OD). The method is evaluated by an experiment on campus networks. Traffic data collected on the backbone network prove that, compared with Generalized Gravity Model(GGM), the method has higher accuracy.
Wireless mobile node can reduce self access data latency and save bandwidth for whole network by caching data, aiming at this characteristic, this paper improves revenue function of Distributed Greedy Algorithm(DGA). It divides the data access frequency into local access frequency and other node access frequency, and researches the application scenes of different weights. NS2 simulation results show that the efficiency of data caching can be enhanced through reasonably selecting weight in different scenes.
According to the too high density of nodes in Wireless Sensor Networks(WSN) and imbalanced energy consumption of nodes, a distributed coverage control algorithm based on energy(DCCABE) is proposed. This algorithm is based on probabilistic coverage model and computes the area coverage probability of nodes to judge the redundancy according to the sequences which are arranged by the remain energy of nodes and changes the redundant nodes into sleeping state. Simulation results show that DCCABE can effectively reduce the redundancy of nodes and prolong the network lifetime.
Single node capability is limited, node in Wireless Sensor Network(WSN) need collaborate with neighbors to complete a task. Aiming at this problem, this paper divides the collaborate task into sensor subtask and computing subtask, and proposes a task collaboration framework based on dynamic coalition and Ant Colony Algorithm(ACO). It chooses the perception node according to the application need, those nodes form an initial dynamic coalition, and assign perception task on the coalition member. When the ratio of perception nodes to the whole nodes is below 32%, the network monitoring performance is best. It adopts self-adaptive ACO to construct data aggregation routing tree, optimizes data transfer path based on strong data relation of a same task, and reduces communication energy consumption.
To extract API-calling behaviors from malware and their variants effectively, this paper proposes an approach to statically detect the obfuscated API-calling behaviors in Windows platform. In this approach, instruction pattern matching is used to recognize the special calling manner. The relationship between the targets of call instructions and the name strings of API functions is analyzed to identify which API function is called actually. Experimental results show that using this approach can improve detection ability of static analysis tools through static analysis.
In the game of attack and defense, semi-distributed P2P botnet becomes the most dominant form of botnet as the applications of P2P are widely used. This paper describes the basic principle of how the attackers create their semi-distributed botnet and the model of its increasing, proposes the “faked honeypot” detection model, which combines honeypot and the flow analysis. When anomalies appear in host network, the programs and services are closed, the host is made to lay aboard to the honeypot, and detected by flow analysis. Experimental result shows that the method can effectively improve the detection rate of semi-distributed P2P botnet.
This paper proposes a forward security proxy multi-signature scheme on the basis of the forward-security idea and proxy multi-signature schemes. The new scheme satisfies security properties of general proxy multi-signature schemes and has forward-security. Under the strong RSA assumption, Diffie-Hellman problem and discrete logarithm problem over finite field, the new scheme has good forward security.
In order to protect the computer from the destruction of the unknown malicious software, this paper carries out the idea of using virtual executing in a simulated system environment to protect computer resources. A lightweight virtual machine is introduced with HOOK to protect critical resources, and a series of API interfaces are provided to make this lightweight virtual machine an open platform available for secondary development. The method successfully identifies new Trojan without feature code.
For the characteristics that Mobile Ad Hoc Network(MANET) has no center and the problems that in general distributed authentications, the number of one-jump certificate nodes is less than the threshold, and CA nodes are unreliable, a certificate service scheme in cluster-based architecture is proposed according to multi-layer-cluster architecture and Feldman’s verifiable secret sharing technology. It achieves the distributed authentication at all levels of cluster head and cluster members. The problem that the number of one-hop neighbor certification nodes is less than threshold value is resolved, which prevents malicious nodes working together to recover the secret key of system effectively, and avoids the interior authentication point broken to the distributional authentication disturbance.
Currently there are few tripartite Authenticated Key Exchange(AKE) protocols which have a formal security proof. Aiming at this problem, this paper investigates the security model for the tripartite AKE protocols. A strong security model named enhanced Canetti- Krawczyk(eCK) model for two-party AKE protocols is proposed. A strong security model for three-party AKE protocols is introduced which generalizes the eCK model for two-party setting, and the insider attack is taken into account. It presents a new tripartite AKE protocol and shows that the protocol is provably secure in the strong security model.
According to the theory of consistency fault diagnosis based on first theory, the max-normal diagnosis theory based on dimidiate- consistent test is proposed. It integrates the kernel theory of consistent fault diagnosis and abductive diagnosis, not only makes up the weakness to construct the conflict set in intelligent fault diagnosis based on model, but also decreases the number of the objects in the consistency test and shrinks the space of diagnosis. Diagnostic accuracy is so greatly improved, the theory has better application. Application example proves the efficiency of this theory.
Eyes location is the priority of iris identification, eye tracking and state recognition. A method for eyes location based on maximum entropy segmentation and skin model is proposed. Image segmentation based on maximum entropy is used to segment the image and candidate regions of eyes are obtained. YCbCr skin model is built as constraint condition of eyes location, and morphological operations are used to obtain precise eyes location. Simulations show that this method can not only precisely and fast locate eyes position, but also adapt to process the image in different background, such as head existing rotation and image itself contain details.
Aiming at the problems that there exists edge vague, the pseudo-edge cannot be removed and algorithm results do not achieve optimal results by virtue of fuzzy reasoning for edge detection algorithms, this paper proposes an improved fuzzy edge detection algorithm. It proposes a fuzzy rule that uses the contrast of gray-scale’s direction to determine the value of edge’s membership and increase the rules to remove false edge and make edge refined. Tests show that the algorithm is better than fuzzy competitive algorithm, Pal King algorithm and other improved fuzzy edge detection algorithms.
Considering the load torque impact on the speed of servo actuator, this paper establishes the model of load torque and proposes a speed circle control algorithm with feedforward fuzzy self-turning Proportional Integral(PI) parameter. It uses feedforward controller to compensate the variational load influence on motor speed. Through dynamic adjustment of control parameter, the controller has good robustness and accuracy when BrushLess DC Motor(BLDCM) characteristics and load torque change. Fuzzy controller is developed by using Digital Signal Processor(DSP) and applying absolute encoder with high accuracy as the feedback of the position loop. Experimental results show that the controller has high robustness, rapid response, small overshoot and high accuracy.
This paper presents a hardware structure-independent weight-based fault injection model for accurate emulation of the radioresistance in the SRAM-based FPGA. Fault injection emulation platform based on Joint Test Action Group(JTAG) boundary scan and dynamic partial reconfiguration is proposed. Experimental results show that fault injection system composed of the software model and the hardware platform has high universal property and is more accurate, efficient and needs lower cost emulation.
On the basis of analyzing the main task processing strategies of multi-task operating system RTX51, this paper designs and implements a data collecting system based on RTX51. According to the executing sequences, the management of data collecting and interface components in the system are divided into five tasks, between which library function calling is used to send signal to coordinate. Under the control of user keys, field data are preserved after converted, and may be displayed and transmitted to supervisory computer memory. The interface files and functions and mechanism for realization of parallel and sequential tasks are presented. Task schedule with the mechanism realizes data collecting.
The approach of modularity, proposed by Newman and Girvan in order to measure the satisfaction with network decomposition, is found not suitable to the evaluation of community structure in networks when the number of links incident to each community differs too much. To resolve this problem, this paper presents a new modularity approach based on the concepts of linking density and cohesion of communities. Through both theoretical and empirical studies, the modularity is applicable to all cases including the networks with communities having nearly the same number of incident links and the networks when the number of links incident to each community differs too much.
As applications programs based on MapReduce model growing up, Hadoop’s performance depends on applications. From applications, this paper analyzes the limitations and shortcomings of Hadoop architecture and storage technology, and proposes a feasible and cost-effective solution to address the problem. It builds a series tests to convince the idea. The solution is multi-level parallel, both on job and I/O levels, which makes full use of disk and network bandwidth, reduces I/O bottlenecks and improves performance.