Recently a Distributed Temporal Protocol Logic has been devised to capture reasoning in the distributed environment of security protocols. Elsewhere we have constructed a proof-based verification framework using distributed temporal protocol logic to verify the authentication property of security protocols. In this paper, we apply our verification framework to a well-known protocol. In particular, we analyze the authentication property of the public-key extension of Kerberos-5 protocol. We demonstrate how we are able to identify a subtle design flaw in the protocol. This results into showing the applicability of our framework as well as demonstrating the ease with which distributed temporal protocol logic can be used to analyze authentication protocols.

In this paper, we propose iterative regularization for image denoising problems, based on the total variation minimizing models proposed by Rudin, Osher, and Fatemi(ROF). Besides, considering the staircase occuring in the process of denoising, we combine the higher order derivatives, and use iterative scheme. The fourth order dual method is used to solve the minimization problems. The numerical experiments show the iterative procedure preserves more details and reduces staircasing. Besides, it can be claimed that the fourth order dual method is more faster and stable than time marching algorithms.

Error bounds between non-stationary binary subdivision curves/surfaces and their control polygons after k-fold subdivision are estimated. The bounds can be evaluated without recursive subdivision. A computational formula of subdivision depth for non-stationary binary subdivision surfaces is also presented. From this formula one can predict the subdivision depth within a user specified error tolerance. Our results not only remove errors in the results of [3] but also contain the generalization of their results.

Random numbers of multivariate nonnormal distribution are strongly requested by the area of theoretic research and application in practice. A new algorithm of generating multivariate nonnormal distribution random numbers is given based on the Copula function, and theoretic analysis suggests that the algorithm is suitable to be feasible. Furthermore, simulation shows that the empirical distribution which is formed by random numbers generating from the proposed algorithm can well approach the original distribution.

Clustering is the main method to analyse the large numbers of data, but when the data’s dimension is higher, the consumed time increases exponentially. We put forward an effective clustering method for high dimensional and large numbers of data, which is based on the sensitive subspace consisting of the data set’s sensitive dimensions. In order to build the sensitive subspace, we first estimate the probability density of each dimension, enhance its optional ability through extracting zero and smoothness processing, then through recognizing the number of the rallying points to gain the sensitive dimensions, and last do the Rival Penalized Competitive Learning (RPCL) clustering in the subspace. Moreover, we detected the red tide of hyper-spectral data using this method, which proved it could effectively get similar results with one-ninth time.

Using a numerical method based on sub-super solution, we will obtain positive solutions for the problem with Dirichlet boundary condition. In particular, we establish

This paper presents an efficient approach to achieve fast and accurate eye states detection in still gray level images with unconstrained background. The structure of eye region is used as a robust cue to find eye pair candidates. Eyes are located by eye verification using SVMs. The eye contour information is used to detect whether eyes are open or closed after locating eyes. The proposed method is robust to deal with illumination changes, moderate rotations, glasses wearing and partial face occlusions. The proposed method is evaluated on the BioID face database and experimental results demonstrate its effectiveness.

This paper aims to improve the performance of the Pinyin-to-Character Conversion system which is the core of Chinese text input method. The ngram model is the current solution to the Pinyin-to- Character Conversion system. This paper enhances the traditional ngram model by relaxing its stationary hypothesis and exploiting the word positional information. The Non-stationary ngram (NS ngram) model is proposed. Several related issues are discussed in detail, including the formal definition, the model implement, the training algorithm and the space complexity of the NS ngram model. Evaluated on the Pinyin-to- Character Conversion task, the NS ngram model outperforms the traditional ngram model significantly with great error rate reductions. Meanwhile, the training algorithm presented in this paper can estimate the parameters in the NS ngram model effectively and efficiently.

Rough set is one of the most useful data mining techniques. How to use rough set to extract rule is the basement of rough set’s application. This paper discuss an algorithm that be used in attribute reduction. To attribute reduction, generally method is based on discernibility matrix or its improvement. But this series methods usually get one reduction, can’t accommodate uncertain information reasoning. We provide a reduction algorithm, which based on reduction pruning. It can calculate all reductions, and suits any uncertain knowledge reasoning. For increase this algorithm’s effective, we present two theorems to make algorithm simplified. We calculate reduction through rough reduction (reduction pruning) and backward elimination two steps. The case illustrates we get the reduction effectively through this algorithm.

This paper aimed to design and implement a new hybrid intelligent technique called expert network. The main purpose of this expert network is to use it as a management tool assist network administrator in his job regarding the monitoring process of network performance. Here, the management task that uses this expert network was how to detect a fault (i.e. fault detection) that may be occurred on a certain part of the network either on hardware or software where this fault can affect on the performance of the computer network as: delay, throughput, utilization. Our proposed expert network depends on using both feed forward Neural Network and CBR. The Multilayer feed forward NN is used to predict the upcoming or the next values of the network major parameters (as load, delay, throughput…etc), then this predicted value will produces a residual signal used for comparing the actual with the desired network parameter. According to the comparison results, the Expert Network will alter the network administrator with certain alarm if the difference between the actual and desired network parameter outside a certain range (threshold value, knowing that this threshold is suggested by the network administrator). Accordingly, the network administrator begins to take some action towards repairing the fault component. Efficiency of the proposed expert-network was examined through investigating the time of detecting a fault. This time was too short relative to the running time of the computer network.

Separation of duty (SOD) is an important characteristic in the role-based access control (RBAC) system. In view of some issues such as various variations of SOD constraints (SODs), ambiguous relations among constraint states, this paper formally defines several typical SODs and analyzes the transition relations among different SODs states. In combination with a delegation case, it goes an exploration and discussion on the SODs state transition issues, and proposes some corresponding solutions.