Oscillatory behavior is a ubiquitous phenomenon in various physical and biological processes. Recently, it has been reported that oscillations of wireless sensor networks infected by virus (WSNIVs) can potentially be deleterious to the security strength of systems and may even cause network congestion and paralysis. Moreover, it has been discovered that latency delays are essential for the function of WSNIVs and can drive instability and periodic oscillations, enhance complexity, and even lead to multistability and chaotic motion. However, the precise roles of such delay during the regulation process are still not completely understood. Here, the primary objective of this paper is to study oscillatory behaviors of WSNIVs with latency delay. In particular, the sufficient conditions for local stability and existence with Hopf bifurcation are obtained. Moreover, we further discuss the properties of Hopf bifurcation by using the normal form and the center manifold theorem. The obtained results show that the latency delay can drive the WSNIVs to be oscillatory even when the network is at a stable state, suggesting that such delay might be a potential hazard to the security of wireless sensor systems. Our findings highlight the importance of considering delays when developing safer and more effective wireless sensor networks. Finally, we test and analyze the above research results through numerical calculation with Matlab and simulation experiments with OPNET, and the conclusions are verified to be correct and effective in experiments.

This paper presents a method to measure what and how deep users can perceive when exploring virtual reality environments using a head mounted display. A preliminary user study was conducted to verify that user gaze behavior has specific differences in immersive virtual reality environments compared with that in conventional, non-immersive virtual reality environments, which are based on a desktop screen. Gathered from the study results for gaze behavior, the users experiencing immersive virtual reality environments are more likely to adjust their head movement to center interesting objects in their vision. Based on this finding, a quantitative method is proposed to measure the user's visual attention in such a virtual reality environment. A user personalized storyboard is designed to capture the user's most regarded views as key frames that can depict the users' exploration experience in immersive virtual reality environments.

Accurate recommendation can effectively bridge sellers and buyers. Because the computation complexity and storage complexity of static data-oriented recommendation algorithms are very high, researchers have recently explored streaming recommendation systems. However, streaming recommendation wastes large quantities of computation resources in quick response and is not suitable for seasonable-dependent situations. Therefore, this paper presents a block incremental recommendation strategy. First, a cascaded trust network construction method is presented, which is realized by using a distrust relationship to purify and predict users' trust relationships. Then, the social regularization is improved by comprehensively considering the cascaded trust relationship, the behavior bias of users and items. Finally, a block-incremental recommendation algorithm called ITDBMF is proposed, which uses the Ebbinghaus forgetting function to decay incremental rating blocks and simultaneously considers incremental social relationships. Experimental results show that the incremental recommendation strategy given in this paper can not only outperform benchmark algorithms in prediction accuracy, but also save storage of remote data and matrix factorization time.

Extracting valuable information from ethnic cultural resources is the key to current data mining research on ethnic cultural resources. The K-means algorithm can effectively process large-scale data sets due to simple and efficient iterative calculations. The uncertainty of the k-value affects the efficiency and accuracy of the algorithm. The particle swarm optimization (PSO) algorithm and global coarse-grained search can quickly determine the k-value of the cluster center, while the retrieval efficiency is low. In order to solve the problem of the initial clustering center of the K-means algorithm and the low efficiency of the PSO algorithm, this paper proposes a Spark-based PSO-k-means algorithm, which primarily introduces ethnic cultural text resources into the Hadoop Distributed File System (HDFS) and then uses Han Language Processing (HanLP) word segmentation. The Term Frequency-Inverse Document Frequency (TF-IDF) algorithm generates the word frequency vector. Finally, the particle swarm optimization algorithm performs initial pre-clustering on the data set, obtains the K-means algorithm cluster center k, and then obtains the final classification result through K-means algorithm cluster analysis. The experimental results show that the clustering accuracy and stability of the PSO-k-means algorithm are better than those of the existing K-means algorithm on serial stand-alone.

Caricature face recognition is an interesting but also difficult task due to the huge exaggeration between two different face modalities, photos, and caricatures. Therefore, we propose a new representation for recognition that is fused by the representation learned from photos, caricatures, and generated faces. Each generated face contains four main facial components. Photos, caricatures, and generated faces are sent to Photo-ResNet, Caricature-ResNet, and Generated-ResNet to learn specific representations. Then, the learned three representations are sent to a fully connected layer. We adopt Softmax loss and Center Loss for training, which can reduce the distance of intra-class. To test the performance of our proposed representation, we build a new dataset for caricature face recognition, which consists of 259 subjects, with 6490 caricatures and 8143 photos. The dataset we build is the biggest available caricature dataset. Several basic methods are used for caricature face recognition. To test the discrimination of our proposed representation, two more experiments are fulfilled, including searching photos according to the selected caricature (CTP) and searching caricatures according to the selected photo (PTC), and our proposed method performs better than other convolutional neural network (CNN)-based representations.

Aiming at the problems of the huge number of malware currently in the big data environment, the insufficient ability of automatic malware analysis available, and the inefficiency of the classification of malicious attributes, in this paper, we propose a Cuckoo-based malware dynamic analysis system that can be extended, analyzed quickly, and has application value. The system proposes a semantic feature model based on deep learning, uses a deep recursive neural network model to describe the multi-layered aggregation relationship of program semantics, and builds a malware semantic aggregation model. The model can automatically complete the acquisition and analysis of behavioural features of unknown program samples and perform attribute discrimination on unknown program samples efficiently and accurately.

As a common and important form, point cloud data exists in computer graphics, especially for 3D visualization. However, with the development of 3D scanning technology, huge data sets have become a main burden in the data processing of point clouds. Therefore, the technology of point cloud compressing is a key content in data pre-processing. This paper provides a new algorithm to compress the point cloud data set. The compressing algorithm can be carried out based on the feature of measure objects. In order to find the data feature, we firstly introduce a point cloud compressing model based on conicoid according to the measure objects. Secondly, for the comparison of the features between the model and the point cloud, we provide a shape operator and a contour operator based on the estimation of geometrical features. Then, according to the value of the shape operator and the contour operator, we provide a matching model. The compressing data algorithm can be created through the matching computation of geometrical features. At last, we use the experiment to prove the feasibility of compressing algorithm, and compare the result of the proposed algorithm and the result of other algorithms in terms of the running time and the compressing effect.

In view of premature convergence and blind searching of the quantum and immune algorithm in the evolution process, this paper proposes two improvements. Firstly, the fitness function is improved by utilizing the mean square error as the fitness function, and the concentration of immune antibodies is introduced to the fitness function to improve the diversity of populations and avoid premature convergence of the algorithm. Secondly, the probability of rotation is adopted to optimize the quantum rotate gate to avoid blind searching and accelerate the convergence of the algorithm. The improved algorithm is adopted to optimize parameters of support vector machines and is applied to network intrusion detection. The experimental results show that the improved algorithm has better optimization effects.

Recommendation systems are widely used in e-commerce websites as they can recommend appropriate movies, songs, books, and other items to users according to users' historical behavior. In traditional collaborative filtering algorithms, users' historical scores are usually used to predict the unknown item rating, while ignoring their textual reviews. Therefore, this paper proposes a park recommendation model based on user reviews and ratings (PRMRR). PRMRR first uses the latent Dirichlet allocation model to extract the statistical distribution of the park features. Secondly, it detects user preference distribution based on park features and user ratings. In order to measure the credibility of user ratings, user rating confidence level is considered to correct user preferences. Thirdly, it uses Kullback-Leibler divergence to calculate the similarity between different users and then predicts the unknown park rating for a specific user. Finally, the proposed algorithm is evaluated on two real park data sets, and the results on two different data sets show that the proposed approach outperforms other traditional approaches. Our recommendation algorithm thus has great potential to improve the quality of park recommendation and effectively handle the data sparsity problem.

The large-scale electromechanical equipment of coal mines has the characteristics of low speed, heavy loads, and complicated operation environment. Existing features, such as shape, color, and texture, are directly used to detect the friction and wear state of large mechanical and electrical equipment in coal mines, and the effect is not satisfactory. In this paper, a multivariate feature extraction algorithm based on maximum wear particles is proposed, and the C-SVC classifier model is constructed based on the extracted features. The simulation results show that compared with SVM (Support Vector Machine) and the decision tree algorithm, the model of C-SVC classifier based on the multiplex feature of the largest block wear particles has better classification accuracy, better generalization ability, and better robustness.

Student performance in higher education is related to many complicated factors and always has uncertainty, so early warning of it is a very difficult issue. In this study, a systematic review was first carried out on student performance prediction and early warning using data mining techniques, including basic data sources, evaluating factors, predicting methods, application tools, and practices. Then, insufficiencies of the related studies were discussed, including incomprehensive source data, inadaptable and unspecialized calculation methods, and lack of integrated methodology systems in practice. Finally, a solution design was proposed, consisting of learning situation big data, a systematic early warning model, and an integrated information support system. Preliminary experiment results showed that it could identify at-risk students in a timely manner and improve the overall efficiency and effectiveness of early warning education management in practice, so it is of both academic and practical significance in promoting the deep integration of information technology and early warning education.

Focused on the issue that text information cannot be fully extracted by the single-channel neural network model, the Dual-Channel Attention Model (DCAM) is proposed for text sentiment analysis. Firstly, text is represented in the form of a matrix using a word vector trained by Word2Vec. Secondly, the matrix is used as input data and sent to Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks for feature extraction. Thirdly, an attention model is introduced to extract important feature information. Finally, the text features are merged, and the classification layer is used to classify the sentiment. The model is evaluated on a Chinese corpus. According to the experimental results, the accuracy of the proposed model can reach 92.7%, which is obviously superior to other single-channel neural network models.

This paper, aiming at the problem of the traditional RD rate control model and inaccurate bit allocation of multi-view high-efficiency video coding (MV-HEVC), proposes a rate control algorithm for MV-HEVC based on the similarity analysis of the views and time-domain complexity of the frame activity. First, the algorithm core uses the similarity analysis of the views to reasonably perform bit allocation among the views. Then, the frame layer and basic unit layer bit allocation and rate control are performed based on the frame rate, target buffer capacity, actual buffer size, and active time domain complexity of the frame. The experimental results show that compared with the fixed bit allocation algorithm applied to views, the proposed method not only effectively controls the bit rate of multi-view video coding with a rate control accuracy of over 99% but also improves the PSNR by 0.38dB.

Node importance ranking of complex networks is of great significance to the study of network robustness. The classical centrality measure degree can reflect the number of neighbors of a node, but it ignores the information between its neighbors. In order to mine the important nodes in the network accurately and efficiently, a method of ranking the node importance of complex networks based on multi-attribute evaluation and node deletion is proposed in this paper. Based on the degree attributes of the target node and its neighbors, this method introduces two attributes, which are the local network density centered on the target node and the assortativity coefficient. It takes into account the characteristics of the scale, tightness, and topology of the local area network where the node and its neighbors are located. This paper conducts deliberate attack experiments on four real networks. Through a comparison between the experimental results of the maximal connected coefficient and network efficiency, our approach is proven to be valid and feasible.

Discovering opinion leaders in social circles is an important issue in social networks. Most existing opinion leader detection methods usually focus on the whole social network. However, the composition of social networks is complicated, as many social circles or communities based on interests exist in social networks. We find that it is hard to find all the opinion leaders of small social circles if we only focus on the whole network. In this work, we propose a method in which we conduct community detection first and then perform influence analysis on the communities to find the opinion leaders of social circles. Most previous overlapping community detection methods are usually time-consuming and cannot output results in acceptable time on a large-scale dataset; therefore, we propose a linear time complexity overlapping community detection method based on topic graph. We calculate degree centrality, betweenness centrality, closeness centrality, and PageRank value of the nodes in each community detected to find opinion leaders. We collect a large-scale dataset from Zhihu and use it to validate our methods. The extensive results demonstrate that our method can produce better results in finding opinion leaders in social circles compared with other methods.

With the advent of MDA, there is an urge to analyze the time performance in real-time systems under various design decisions at the very early stages of design. With the wide application of customized real-time operating system (RTOS) based on a microkernel, we propose a time-oriented modeling and analysis approach for real-time systems based on RTOS at the early stages of design. According to the commonality and variability in the system, a modeling approach for analyzing the time under variable design decisions is presented. These design decisions include various hardware environment, user-level services adopted in RTOS, and the task settings. In the analysis approach, a timing tree with the operating and timing rules is defined and used based on the time annotations of the basic system call of RTOS and worst-case execution time (WCET) of the functional block in a task to analyze the execution time. The modeling and analysis approach proposed is capable of analyzing new decisions without any changes in the model, which is helpful to find the best design decision to improve the real-time in the system. Finally, a real-life aircraft landing control system is taken as an example to evaluate this approach.

With the development of social economy, science and technology, and the popularity of mobile Internet, human society has gradually entered the information age. The rapid growth and disorderly distribution of information have posed challenges to English learners who conduct independent and individualized learning anytime and anywhere. Reading plays an important role in language learning. Traditionally, it is difficult for the English reading method based on chapters to meet the fragmented learning needs of learners who are often in a mobile environment. Through the literature research, this paper designed the five-dimensional feature model of learners and the three-dimensional feature model of English reading resources in the fragmented learning environment. Combined with the machine learning algorithm ID3, the decision tree classification model of English fragmented reading was constructed to push resources, adapt to the individual needs of the learners, and then specifically enhance the different aspects of the learner's reading ability, thereby improving the pass rate of the College English Test-Band Four.

The remaining useful life is key to the decision-making of machinery maintenance. The online prediction of remaining useful life has become a very urgent need for mechanical equipment with high reliability requirements. The aim of this paper is to provide a simple and effective method for predicting the remaining life of the machine under the condition of small sample. The Kolmogorov-Smirnov test theory is used to extract the health state feature of the machine. Based on the Long and Short Term Memory (LSTM) theory, an online method of remaining useful life prediction is proposed. The bearing life vibration data verification shows that the Kolmogorov-Smirnov distance is sensitive to the development and expansion of the defects. Furthermore, the proposed method of remaining useful life prediction based on LSTM theory has high prediction accuracy. The technician can then use this method to take appropriate maintenance operations.

At present, the sharing and dissemination of cultural resources of minorities in China remain at the early information service stage based on search engine and database query, with slow content updates, closed structure, independence from other databases, and disconnection of content, and they are far from meeting the actual needs of the sharing and dissemination of cultural resources of ethnic minorities. Therefore, aiming at the heterogeneity problem in the sharing and service of minority cultural resources and based on theories and methods such as domain ontology and Map Reduce, this paper first constructs a multi-source heterogeneous integrated model of massive minority cultural resources, Then, an example of Wa nationality is applied to construct the resource domain ontology of ethnic minorities and expand the domain. Finally, on the basis of semantic distance, a method of weighted comprehensive semantic similarity calculation is proposed and verified. The experimental results show that the similarity of the parent node and each child node in Wa hierarchical tree is different, and the similarity result is more reasonable than the original method.

Coloring animation sketches has always been a complex and interesting task, but as the sketch is the first part of animation creation that neither presents gray value nor presents semantic information, the lack of real animation sketches is the biggest difficulty in current model training. It is also usually expensive to collect such data. In recent years, some methods based on generative adversarial networks (GANs) have achieved great success. They can generate colorized anime illustration on given sketches. Many existing sketch coloring tools are based on this supervised learning method, but the marking of data is particularly important for supervised learning, and much time is spent on the marking of data. To address these challenges, we propose a novel approach for unsupervised learning based on U-net and periodic consistent confrontation. Specifically, we combine the periodic consistent antagonism framework with the U-net structure and residual network, enabling us to robustly train a deep network to make the resulting images more natural and realistic. We also adopted some special data generation methods, so that our model can not only color anime sketches but also extract line drafts from colored pictures. By comparing the mainstream models of supervised learning, we show that the image processed by the proposed method can achieve a similar effect.

LiDAR (Light Detection and Ranging) is a laser ranging technology that provides an efficient and convenient way to obtain the original data from DSM (Demand Side Management). The basic task of LiDAR is to separate the high quality DTM (Digital Terrain Model) from the DSM, and the accuracy and quality of the generated image are determined by the different filtering and interpolation algorithms. According to this, this paper presents a filtering algorithm based on the optimal threshold segmenting optimized by the erosion operation (OTS-EO) to improve the problem that the manually-set-height difference threshold is empirically affected. In order to overcome the deficiency of the distance-based IDP (Inverse Distance to a Power) interpolation algorithm, an interpolation algorithm based on elevation and distance weighting is proposed to obtain the DSM to be further filtered. In this paper, the original laser point cloud data near the Xinyan rode in Beijing is taken as an example, and the data is processed by the algorithm based on threshold segmentation and mathematical morphology (TSMM) to extract the DTM. Finally, the 3D visualization of DTM is realized by the program based on MFC and OpenGL. The experimental data and practices in engineering show that the TSMM algorithm can successfully separate and display the surface points and surface features and extract the DTM close to the real ground to provide the foundation for further research.

With the advent of the era of big data, the problem of information overload has become particularly serious. The recommendation system can provide personalized recommendation services for users by analyzing users' basic information and users' behavior information. How to push information accurately and efficiently has become an urgent issue in the era of big data. Based on the Alternating Least Squares (ALS) collaborative filtering recommendation algorithm, this paper reduces the loss of the invisible factor item attribute information by merging the similarity of the item on the loss function. At the same time, the cold start strategy is introduced into the model to complete the recommendation. The algorithm is implemented on the Spark distributed platform and single node by using the Movie Lens dataset published by the GroupLens Lab. The experiment results show that the proposed recommendation algorithm can preferably alleviate the data sparsity problem compared with the traditional recommendation algorithm. Moreover, the algorithm improves the accuracy of recommendation and the efficiency of calculation.

An object usually belongs to multiple concepts, but some concepts can be judged directly while other concepts need to be inferred indirectly. To learn some arithmetic concepts from positive integer number sets, we address an arithmetic concept Bayesian network (ACBN) model by taking advantage of Bayesian networks. Specifically, we first give an ACBN model to represent the arithmetic concept knowledge and their direct relationships, and then we design an ACBN model learning algorithm based on domain knowledge. Furthermore, to infer indirectly some arithmetic concepts, we design the learning method of evidence concepts based on the idea of k-nearest neighbors, and then we propose the inference algorithm of the ACBN model. Finally, the experimental results demonstrate that the ACBN model can effectively learn some daily arithmetic concepts.

In a wireless sensor network, authentication of multiple nodes can effectively reduce the overhead of node communication and improve the security of wireless sensor network operation. However, when the node was authenticated by existing methods, the positional information of the node was easily exposed, resulting in a high calculation repetition rate and increased consumption of node communication. A random unite authentication method based on RSA and trusted nodes was proposed. The multi-node authentication model of the RSA method was used to effectively authenticate other nodes in the wireless sensor network. The node trust degree was introduced, and the node authentication key was updated by the identity-based and bilinear theory. After the trust reputation management based on the Beta distribution node was used to calculate the trust degree of the node, it was verified whether the trust degree of the communication node was trustworthy to the node; the verification result was identified. Finally, the method of combining symmetric cryptography with node information authentication code was used to implement trusted node authentication in wireless sensor networks. Experiments showed that the authentication node results of this mechanism were stable, which overcame the shortcomings of low stability of multi-node authentication results in the current method.

The application of robots plays an important role in the development of intelligent production and life. At present, most robots avoid the obstacle in the process of robot operation through a single ultrasonic ranging, and it cannot guarantee the accuracy of the obstacle avoidance of robots. In this paper, an intelligent distance measurement method of robot obstacle avoidance in a cloud computing environment is designed and studied. Based on the DSP and ultrasonic global positioning system, a multi-channel ultrasonic transmitter/receiver module is adopted to design an autonomous obstacle avoidance control system based on ultrasonic waves and a new fuzzy reasoning method is proposed to realize the function of intelligent distance measurement of robot obstacle avoidance in the cloud computing environment. The simulation and field test for the intelligent distance measurement system of the obstacle avoidance is carried out by Visual C ++ visual programming software. The reliability and feasibility of the system are verified, which provides a wider space for the research and development of the robot.

The confidentiality of data is a difficult problem in a cloud computing environment. The DAC technique (Dynamic Access Control) based on encrypted data is an important way to solve this problem. In the current access control techniques based on encrypted data, the high-security requirement of data and frequent policy updates lead to the high cost of owner right update which seriously restricts the flexibility of access control. A DCA method based on CACDP encrypted data is proposed. The selective encryption model is built. In the model, a key derivation diagram is generated to distribute the key. In the case of ensuring the confidentiality of cloud computing access control, the key is less in the system. The proposed CACDP scheme includes the key management mechanism based on the binary Trie tree. Based on this, the ELGamal-based proxy re-encryption mechanism and double layer encryption strategy are used to transfer the partial spending of the key and data update to the cloud to ease the DO authority management burden and increase the efficiency of DO. Then, the DCA method of encrypted data in cloud computing environment is researched. Experimental results show that our proposed method can effectively improve the flexibility of encrypted data access control.

In order to strengthen the logistics distribution ability at the end of e-commerce, it is necessary to study the co-delivery collaborative delivery method of e-commerce. When the current delivery method is used to deliver the courier at the end of the e-commerce, the distribution takes a long time to meet the distribution requirements of the user, and there is a problem of low distribution efficiency and low customer satisfaction. On the basis of the big data platform, an e-commerce end express co-delivery distribution method is proposed. By calculating the inventory cost, fixed investment cost, storage cost and operation cost, the total cost of the distribution center is coordinated, and the lowest total cost is selected as the distribution center. The distribution time range is analyzed by the concept of time window to obtain a penalty function. According to the penalty function, the path optimization model of e-commerce end express delivery is established. According to the path optimization model, the optimal route of collaborative delivery is obtained, and the coordinated delivery of the e-commerce end express is completed. The experimental results show that the proposed method has high distribution efficiency and high customer satisfaction.

In order to reduce the labor intensity of workers in automatic production lines handling material, an Automated Guided Vehicle collaboration system (AGV) is designed in this paper. On the basis of analyzing the differential steering principle, the control strategy of the AGV tracking and the automatic tracking control are designed based on the fuzzy algorithm and PID algorithm respectively, so as to realize the control of PC for the transport of goods. Additionally, the PC control software is specially designed for the system. This system is tested on the simulation and experiment environment and the results show that the AGV has the advantages of high guidance precision, following stability and high reliability. The fleet operation can effectively improve the work efficiency and reduce labor costs.

With the rapid development of communication technology, digital signal processing and other technologies, wireless communication environment is becoming more and more complex. Communication signals with different frequencies and modulated modes are usually scattered over a wide frequency band. In this paper, an improved algorithm based on wavelet transform and fractal theory is proposed. To improve the traditional fractal theory, wavelet transform is applied to the modulation signal, and then four fractal dimensions (Fractal box dimension, Petrosian fractal dimension, Katz fractal dimension and Sevcik fractal dimension) are used to extract the features. Through the simulation of the six modulation signals generated by Matlab, it can be seen that the recognition rate of the proposed method reaches 90% at the SNR of 2dB. Moreover, by comparing the method of this paper with the short-time Fourier transform and the fractional Fourier transform, we can find that the recognition rate of this method is 3% ~ 10% higher than the two comparison methods. It can be seen that the proposed method can effectively identify different signals in the case of low SNR.

Specific emitter identification is the process of identifying or discriminating different emitters based on the radio frequency fingerprints extracted from the received signal. Due to the inherent non-linearities of the power amplifiers of emitters, these features are extracted from the power amplifiers for specific emitter identification. In this paper, we take the 433MHz power amplifier module BLT54A as the research object, eight circuit modules with the same specifications and batches were tested. The experimental contents include S parameters, AM-AM curve, AM-PM curve, input power and output power curve, AM-AM curve when bias voltage is different, AM-PM curve and input power and output power curve. We extract the energy characteristics of the HHT spectrum of the amplifier output signal and recognize it by classifier. The results show that the eight amplifiers can reach 75%. Meanwhile, we evaluate performance of specific emitter identification in several aspects including different signal sample length and different SNR.

The identification of target identity attributes and its behavioral prediction are important means for providing command and decision support in modern warfare. This paper analyzes the key steps in the process of target recognition and behavior prediction and proposes a target recognition and behavior prediction model based on the Bayesian network. In the simulation example, by integrating the sensor information and combining expert knowledge, the model can effectively and accurately conduct battlefield situational awareness. Combined with the background of big data, this paper introduces the distributed processing system Hadoop, and prospects its application in target recognition and behavior prediction.

In order to increase the security of sensitive information, we need to research the security storage of sensitive information in CCDC (Cloud Computing Data Center). The traditional algorithm cannot guarantee the security of sensitive information when the hacker attacks the data center. Therefore, this paper proposes the three-dimensional CCDC sensitive information security storage algorithm. This algorithm uses the feature combination to filter the sensitive information and uses the sensitive information phase extraction technology to encrypt the sensitive information that was screened. This algorithm also uses three dimensional storage principle for the security storage of sensitive information. Experimental results show that this algorithm can effectively enhance the security of the sensitive information of CCDC.

The WiFi-based fingerprinting technique is widely adopted for indoor positioning due to its cost-effectiveness compared to other infrastructure-based positioning methods. However, the WiFi-based technique still faces the problem of device diversity in the application of an indoor positioning system. Previous studies have faced two main challenges. One is the curse of computational dimensionality in online positioning, while the other is the issue of low positioning accuracy in real applications. In this paper, we propose to normalize the observable Access Point (AP) signal strength to eliminate the influence of device diversity and avoid a dimension disaster. Experimental results show that our algorithm based on the normalization Received Signals Strength (RSS) not only solves the problem of device diversity but also outperforms three other baseline methods.

According to the provisions of the SOLAS convention, more than 600,000 ships in the world have installed automatic identification systems, which can produce more than 3 billion lines of AIS message data every day. The large amount of AIS message data provides many applications. At the same time, the design of efficient decoding of AIS messages has become a new problem. The former AIS message decoding methods mostly use online real-time decoding and decode the messages line by line according to the order of messages. These methods are slow and inefficient when decoding large amounts of AIS messages, and they cannot use multi-core CPU computing power to the extreme. In this paper, the decoding process is decomposed into several steps such as data reading, single-line message filtering, multi-line message filtering, single-line message binary transcoding, multi-line message binary transcoding, binary code decoding, and preservation results. Because multi-line message filtering and multi-line message binary transcoding are difficult to perform in parallel, these two steps are performed in serial, while the other steps are run in parallel. For binary code decoding, which consumes the largest computing resources in the process, the data is blocked first, and then decoding tasks are created and run in parallel. This method makes full use of the parallel computing capability of multi-core CPU and achieves high speed in the running test.

Binary code analysis is the main method for malware analysis. In this paper, the analysis is started by identifying return instructions to disassemble binary code. The return instruction identification problem is converted into a binary classification problem is a byte in binary code the first byte of a return instruction? The 32 bytes around a byte in binary code are considered the feature of the byte. A multilayer perceptron is employed to build the classification model. Then, the model is trained with 1,383 binaries from Windows XP SP3. The evaluation results on several open sources show that our approach is feasible and has high accuracy.