Aiming at the problem that the semantic ambiguity caused by the lack of context information in Chinese short texts results in feature sparsity, a text classification model combing Convolutional Neural Network and Multi-Head self-Attention mechanism (CNN-MHA) was proposed. Firstly, the existing Bidirectional Encoder Representations from Transformers (BERT) pre-training language model was used to format the sentence-level short texts in the form of character-level vectors. Secondly, in order to reduce the noise, the Multi-Head self-Attention mechanism (MHA) was used to learn the word dependence inside the text sequence and generate the hidden layer vector with global semantic information. Then, the hidden layer vector was input into the Convolutional Neural Network (CNN) to generate the text classification feature vector. In order to improve the optimization effect of classification, the output of convolutional layer was fused with the sentence features extracted by BERT model, and then inputted to the classifier for re-classification. Finally, the CNN-MHA model was compared with TextCNN model, BERT model and TextRCNN model respectively. Experimental results show that, the F1 performance of the improved model is increased by 3.99%, 0.76% and 2.89% respectively compared to those of the comparison models on SogouCS dataset, which proves the effectiveness of the improved model.

Analysis and research on interbank market based on multi-layer network structure is conducive to avoiding or weakening the risk impact on financial market. Based on test data simulated by credit lending business scenario, combined with the multi-layer network structure and complex network analysis method of interbank market, the important nodes in interbank market were judged and identified from different angles, meanwhile Jaccard similarity coefficient between the layers and inter-institution Pearson similarity coefficient were calculated and the infectousness of risk contagion of interbank market was measured from macroscopic and microscopic perspectives. The experimental results show that large-scale state-owned financial institutions such as Bank of China and China Development Bank are of high importance in the system, and the greater the similarity between institutions, the greater the infectiousness of risk contagion. Therefore, by calculating the important node measure index in the network layer, comprehensive and complete analysis of the risk contagion of the entire system can help the regulators to achieve accurate monitoring of important institutions in the system. At the same time, from the perspectives of inter-layer analysis and intra-layer analysis, comprehensive measurement of the infectious degree of risk contagion between institutions after financial shock provides policy advice to regulators.

In order to address the problem that the mutation in the current fuzzing has certain blindness and the samples generated by the mutation mostly pass through the same high-frequency paths, a binary fuzzing method based on light-weight program analysis technology was proposed and implemented. Firstly, the target binary program was statically analyzed to filter out the comparison instructions which hinder the sample files from penetrating deeply into the program during the fuzzing process. Secondly, the target binary program was instrumented to obtain the specific values of the operands in the comparison instructions, according to which the real-time comparison progress information for each comparison instruction was established, and the importance of each sample was measured according to the comparison progress information. Thirdly, the real-time path coverage information in the fuzzing process was used to increase the probability that the samples passing through rare paths were selected to be mutated. Finally, the input files were directed and mutated by the comparison progress information combining with a heuristic strategy to improve the efficiency of generating valid inputs that could bypass the comparison checks in the program. The experimental results show that the proposed method is better than the current binary fuzzing tool AFL-Dyninst both in finding crashes and discovering new paths.

The deobfuscation result of deobfuscation framework Miasm is a picture, which cannot be decompiled to recovery program source code. After deep research on the obfuscation strategy of Obfuscator Low Level Virtual Machine (OLLVM) and Miasm deobfuscation idea, a general OLLVM automatic deobfuscation framework based on symbolic execution was proposed and implemented. Firstly, the basic block identification algorithm was used to find useful basic blocks and useless blocks in the obfuscated program. Secondly, the symbolic execution technology was used to determine the topological relations among useful blocks. Then, the instruction repairment was directly applied to the assembly code of basic blocks. Finally, an executable file after deobfuscation was obtained. The experimental results show that, under the premise of guaranteeing the deobfuscation time as little as possible, the code similarity between the deobfuscation program and the non-obfuscated source program is 96.7%. The proposed framework can realize the OLLVM deobfuscation of the C/C ++ files under the x86 architecture very well.

The Iterative Closest Point (ICP) algorithm requires two point clouds to have a good initialization to start, otherwise the algorithm may easily get trapped into local optimum. In order to solve the problem, a novel translation domain estimating based global point cloud registration algorithm was proposed. The translation domain was estimated according to axis-aligned bounding box of calculating the defuzzification principal point clouds of data and model point clouds. With the estimated translation domain and [-π, π] ³ rotation domain, an improved globally optimal ICP was used to register for global searching. The proposed algorithm could estimate translation domain adaptively and register globally according to the point clouds for registration. The process of registration did not need to calculate the feature information of point clouds and was efficient for any initialization with less setting parameters. The experimental results show that the proposed algorithm can get accurate registration results of global optimization automatically, and also improve the efficiency of global registration.

With the development of e-commerce, online shopping Web sites provide reviews for helping a customer to make the best choice. However, the number of reviews is huge, and the content of reviews is typically redundant and non-standard. Thus, it is difficult for users to go through all reviews in a short time and find the distinguishing characteristics of a product from the reviews. To resolve this problem, a method to mine top- k emerging patterns was proposed and applied to mining reviews of different products. Based on the proposed method, a prototype, called ReviewScope, was designed and implemented. ReviewScope can find significant comments of certain goods as decision basis, and provide visualization results. The case study on real world data set of JD.com demonstrates that ReviewScope is effective, flexible and user-friendly.

According to the poor applicability and poor prediction accuracy fluctuation of the existing Software Reliability Growth Model (SRGM), this paper proposed a model based on Fuzzy Neural Network (FNN) which was connected with self-Adaptive Step Cuckoo Search (ASCS) algorithm, the weights and thresholds of the FNN were optimized by ASCS algorithm, then the FNN was used to establish SRGM. Software defect data were used in the FNNs training process, the weights and thresholds of FNN were adjusted by ASCS, the accuracy of prediction process was improved correspondingly, at the same time, in order to reduce the fluctuation of prediction by FNN, averaging method was used to deal with predicted results. Based on those, SRGM was established by self-Adaptive Step Cuckoo Search algorithm—Fuzzy Neural Network (ASCS-FNN). According to 3 groups of software defect data, taking Average Error (AE) and Sum of Squared Error (SSE) as measurements, the SRGMs one-step forward predictive ability established by ASCS-FNN was compared with the SRGMs one-step forward predictive ability established by Simulated Annealing—Dynamic Fuzzy Neural Network (SA-DFNN), FNN and Back Propagation Network (BPN). The simulation results confirm that, the SRGM based on ASCS-FNN relative to the SRGM based on SA-DFNN, FNN and BPN, the mean of Relative Increase (RI) of prediction accuracy rate for RI (AE) is -1.48%, 54.8%, 33.8%, and the mean of Relative Increase (RI) of prediction accuracy rate for RI (SSE) is 14.4%, 76%, 35.9%. The prediction of SRGM established by ASCS-FNN is more steadily than the prediction of SRGM established by FNN and BPN, and the net structure of ASCS-FNN is much simpler than the net structure of SA-DFNN, so the SRGM established by ASCS-FNN has high prediction accuracy, prediction stability, and some adaptability.

Based on the framework of Technology Acceptance Model (TAM), this paper built the users' acceptance behaviour theoretical model of hedonic information system by theory analysis and adding new variables, in order to reveal the users' acceptance behaviour rules of mobile reading. This paper collected data by questionaire, analyzed data, tested hypothesis and modified model with SPSS and AMOS. Results reveal that electronic service quality has positive relationship with perceived usefulness and perceived ease of use, which makes up the defect that TAMs do not pay enough attention to the external factors; users' flow experience has positive relationship with attitude and intention, which breaking through TAM's rational behaviour premise. It expands TAM by introducing the emotional variables that reflect users' emotion on hedonic information system.