Aiming at the problems of lack of Industrial Control System （ICS） data and poor detection of unknown attacks by industrial control intrusion detection systems， an unknown attack intrusion detection method for industrial control systems based on Generative Adversarial Transfer Learning network （GATL） was proposed. Firstly， causal inference and cross-domain feature mapping relations were introduced to reconstruct the data to improve its understandability and reliability. Secondly， due to the data imbalance between source domain and target domain， domain confusion-based conditional Generative Adversarial Network （GAN） was used to increase the size and diversity of the target domain dataset. Finally， the differences and commonalities of the data were fused through domain adversarial transfer learning to improve the detection and generalization capabilities of the industrial control intrusion detection model for unknown attacks in the target domain. The experimental results show that on the standard dataset of industrial control network， GATL has an average F1-score of 81.59% in detecting unknown attacks in the target domain while maintaining a high detection rate of known attacks， which is 63.21 and 64.04 percentage points higher than the average F1-score of Dynamic Adversarial Adaptation Network （DAAN） and Information-enhanced Adversarial Domain Adaptation （IADA） method， respectively.

Private Set Intersection （PSI） is an important solution for privacy information sharing. A fair multi-party PSI protocol based on cloud server was proposed for the unfairness caused by the existing protocols in which the parties involved do not have simultaneous access to the calculation results. Firstly， the storage of a sub-share of the private information in Garbled Bloom Filter （GBF） was accomplished by using hash mapping. Secondly， in order to avoid the leakage of the index value of each party’s set element during the interaction， combined with Oblivious Transfer （OT） technique， the share replacement of the stored information was realized. Finally， the bit-by-bit calculation was performed by the cloud server， and the results were returned to each party at the same time to ensure the fairness of each party’s access to the results. The correctness and security analysis of the protocol shows that the proposed protocol can achieve the fairness of the parties in obtaining the intersection results， and can resist the collusion of parties with the cloud server. The performance analysis shows that both of the computational complexity and the communication complexity of the proposed protocol are independent of the total number of elements contained in the set of participants. Under the same conditions， compared with Multi-party PSI protocol （MPSI）， practical multiparty maliciously-secure PSI protocol （PSImple） and Private Intersection Sum algorithm （PI-Sum）， the proposed protocol has less storage overhead， communication overhead and running time.

In automated industrial scenarios， the amount of time series log data generated by a large number of industrial devices has exploded， and the demand for access to time series data in business scenarios has further increased. Although HBase， a distributed column family database， can store industrial time series big data， the existing strategies cannot meet the specific access requirements of industrial time series data well because the correlation between data and access behavior characteristics in specific business scenarios is not considered. In view of the above problem， based on the distributed storage system HBase， and using the correlation between data and access behavior characteristics in industrial scenarios， a distributed storage performance optimization strategy for massive industrial time series data was proposed. Aiming at the load tilt problem caused by characteristics of industrial time series data， a load balancing optimization strategy based on hot and cold data partition and access behavior classification was proposed. The data were classified into cold and hot ones by using a Logistic Regression （LR） model， and the hot data were distributed and stored in different nodes. In addition， in order to further reduce the cross-node communication overhead in storage cluster and improve the query efficiency of the high-dimensional index of industrial time series data， a strategy of putting the index and main data into a same Region was proposed. By designing the index RowKey field and splicing rules， the index was stored with its corresponding main data in the same Region. Experimental results on real industrial time series data show that the data load distribution tilt degree is reduced by 28.5% and the query efficiency is improved by 27.7% after introducing the optimization strategy， demonstrating the proposed strategy can mine access patterns for specific time series data effectively， distribute load reasonably， reduce data access overhead， and meet access requirements for specific time series big data.

Aiming at the problem that the existing adversarial example generation methods require a lot of queries to the target model， which leads to poor attack effects， a Text Adversarial Examples Generation Method based on BERT （Bidirectional Encoder Representations from Transformers） model （TAEGM） was proposed. Firstly， the attention mechanism was adopted to locate the keywords that significantly influence the classification results without query of the target model. Secondly， word-level perturbation of keywords was performed by BERT model to generate candidate adversarial examples. Finally， the candidate examples were clustered， and the adversarial examples were selected from the clusters that have more influence on the classification results. Experimental results on Yelp Reviews， AG News， and IMDB Review datasets show that compared to the suboptimal adversarial example generation method CLARE （ContextuaLized AdversaRial Example generation model） on Success Rate （SR）， TAEGM can reduce the Query Counts （QC） to the target model by 62.3% and time consumption by 68.6% averagely while ensuring the SR of adversarial attacks. Based on the above， further experimental results verify that the adversarial examples generated by TAEGM not only have good transferability， but also improve the robustness of the model through adversarial training.

The sugar manufacturing enterprises use traditional algorithm to forcast the raw sugar demand, which ignors the influence of time factors and the industry characteristics, resulting in low accuracy. To address this problem, combining with the periodic characteristics of the supply and demand of raw materials of refining sugar,a temporal feature-correlated raw sugar demand forecast model based on improved Elman Neural Network with Modified Cuckoo Search(MCS) optimization was proposed, namely TMCS-ENN. Firstly, an adaptive learning rate formula was proposed to optimize Elman Neural Network (ENN). Secondly, the adaptive parasitic failure probability and adaptive step-length control variable formula were introduced to obtain MCS algorithm to optimize the weight and threshold of ENN, which effectively improved the local search ability of the model and avoided local optimum. Finally, combining time correlation and hysteresis of raw material purchase of sugar manufacturing enterprise, the data slices were designed based on week granularity, and the ENN was trained with festivals and holidays as important features to obtain TMCS-ENN. Experimental results show that, with week as time granularity, the forecasting accuracy of the proposed TMCS-ENN forecasting model reaches 93. 89%. It can be seen that TMCS-ENN can meet the forecast accuracy demand of sugar manufacturing enterprises and effectively improve their production efficiency.

Aiming at the problem that the existing human skeleton-based action recognition algorithms cannot fully explore the temporal and spatial characteristics of motion, a human skeleton-based action recognition algorithm based on Spatiotemporal Attention Graph Convolutional Network (STA-GCN) model was proposed, which consisted of spatial attention mechanism and temporal attention mechanism. The spatial attention mechanism used the instantaneous motion information of the optical flow features to locate the spatial regions with significant motion on the one hand, and introduced the global average pooling and auxiliary classification loss during the training process to enable the model to focus on the non-motion regions with discriminability ability on the other hand. While the temporal attention mechanism automatically extracted the discriminative time-domain segments from the long-term complex video. Both of spatial and temporal attention mechanisms were integrated into a unified Graph Convolution Network (GCN) framework to enable the end-to-end training. Experimental results on Kinetics and NTU RGB+D datasets show that the proposed algorithm based on STA-GCN has strong robustness and stability, and compared with the benchmark algorithm based on Spatial Temporal Graph Convolutional Network (ST-GCN) model, the Top-1 and Top-5 on Kinetics are improved by 5.0 and 4.5 percentage points, respectively, and the Top-1 on CS and CV of NTU RGB+D dataset are also improved by 6.2 and 6.7 percentage points, respectively; it also outperforms the current State-Of-the-Art (SOA) methods in action recognition, such as Res-TCN (Residue Temporal Convolutional Network), STA-LSTM, and AS-GCN (Actional-Structural Graph Convolutional Network). The results indicate that the proposed algorithm can better meet the practical application requirements of human action recognition.

Concerning the problems that the traditional Convolutional Neural Network (CNN) ignores the context and semantic information of words and loses a lot of feature information in maximal pooling processing, the traditional Recurrent Neural Network (RNN) has information memory loss and vanishing gradient, and both CNN and RNN ignore the importance of words to sentence meaning, a model based on parallel hybrid network and attention mechanism was proposed. First, the text was vectorized with Glove. After that, the CNN and the bidirectional threshold recurrent neural network were respectively used to extract text features with different characteristics through the embedding layer. Then, the features extracted by two networks were fused. And the attention mechanism was introduced to judge the importance of different words to the meaning of sentence. Multiple sets of comparative experiments were performed on the English corpus of IMDB. The experimental results show that the accuracy of the proposed model in text classification reaches 91.46% and F1-Measure reaches 91.36%.

The existing deep learning based infrared and visible image fusion methods have too many parameters and require large amounts of computing resources and memory. These methods cannot meet the deployment demand of resource constrained edge devices such as cell phones and embedded devices. In order to address these problems, a light-weight image fusion method based on SqueezeNet was proposed. SqueezeNet was used to extract image features, then the weight map was obtained by these features, and the weighted fusion was performed, finally the fused image was generated. By comparing with the ResNet50 method, it is found that the proposed method compresses the model size and network parameter amount to 1/21 and 1/204 respectively, and improves the running speed to 5 times while maintaining the quality of fused images. The experimental results demonstrate that the proposed method has better fusion effect compared to existing traditional methods as well as reduces the size of fusion model and accelerates the fusion speed.

Movidius neural computing stick is a USB-based deep learning inference tool and a stand-alone artificial intelligence accelerator that provides dedicated deep neural network acceleration for a wide range of mobile and embedded vision devices. For the embedded application of deep learning, a near real-time pedestrian target detection method based on Movidius neural computing stick was realized. Firstly, the model size and calculation were adapted to the requirements of the embedded device by improving the RefineDet target detection network structure. Then, the model was retrained on the pedestrian detection dataset and deployed on the Raspberry Pi equipped with Movidius neural computing stick. Finally, the model was tested in the actual environment, and the algorithm achieved an average processing speed of 4 frames per second. Experimental results show that based on Movidius neural computing stick, the near real-time pedestrian detection task can be completed on the Raspberry Pi with limited computing resources.

Aiming at the problems of insufficient use of semantic information and the poor summarization precision in the current generated text summarization method, a text summarization method was proposed based on dual encoder. Firstly, the dual encoder was used to provide richer semantic information for Sequence to Sequence (Seq2Seq) architecture. And the attention mechanism with dual channel semantics and the decoder with empirical distribution were optimized. Then, position embedding and word embedding were merged in word embedding technology, and Term Frequency-Inverse Document Frequency (TF-IDF), Part Of Speech (POS), key Score (Soc) were added to word embedding, as a result, the word embedding dimension was optimized. The proposed method aims to optimize the traditional sequence mapping of Seq2Seq and word feature representation, enhance the model's semantic understanding, and improve the quality of the summarization. The experimental results show that the proposed method has the performance improved in the Rouge evaluation system by 10 to 13 percentage points compared with traditional Recurrent Neural Network method with attention (RNN+atten) and Multi-layer Bidirectional Recurrent Neural Network method with attention (Bi-MulRNN+atten). It can be seen that the proposed method has more accurate semantic understanding of text summarization and the generation effect better, and has a better application prospect.

Concerning the problem of traditional Particle Swarm Optimization (PSO) such as slow convergence and being easy falling into local extremum, a Dynamic self-Adaptive Chaotic Quantum-behaved PSO (DACQPSO) was proposed by studying the relationship between population diversity and the evolution of PSO. The population-distribution-entropy was introduced into the evolutionary control of the particle swarm in this algorithm. Based on the Sigmoid function model, the method of calculating the contraction-expansion coefficient of the Quantum-behaved PSO (QPSO) was given. The average-distance-amongst-points was taken as the criterion of chaotic search to carry out a chaotic perturbation. The DACQPSO algorithm was applied to the coverage optimization of Wireless Sensor Network (WSN), and the simulation analysis was carried out. Experimental results show that compared with Standard PSO (SPSO), QPSO and Chaotic Quantum-behaved PSO (CQPSO), the DACQPSO algorithm improves the coverage rate by 3.3501%, 2.6502% and 1.9000% respectively. DACQPSO algorithm improves the coverage performance of WSN, and has better coverage optimization effect than other algorithms.

Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) are widely used in natural language processing, but the natural language has a certain dependence on the structure, only relying on CNN for text classification will ignore the contextual meaning of words, and there is a problem of gradient disappearance or gradient dispersion in the traditional RNN, which limits the accuracy of text classification. A feature fusion model for CNN and Bidirectional Long Short-Term Memory (BiLSTM) was presented. Local features of text were extracted by CNN and global features related to text were extracted by BiLSTM network. The features extracted by the two complementary models were merged to solve the problem of ignoring the contextual semantic and grammatical information of words in a single CNN model, and the fusion model also effectively avoided the problem of gradient disappearance or gradient dispersion in traditional RNN. The experimental results on two kinds of datasets show that the proposed fusion feature model can effectively improve the accuracy of text classification.

Concerning the high complexity of the traditional frequency estimation algorithm, a new frequency offset estimation algorithm of Time Domain Synchronous Orthogonal Frequency Division Multiplexing (TDS-OFDM) with low complexity for power line communication was proposed. Firstly, the characteristics of power line network were analyzed, and a frame head was constructed with three equal-length cyclic Pseudo-random Noise (PN) sequences. Secondly, the frame head and body were based on Binary Phase Shift Keying (BPSK) and Quadrature Amplitude Modulation (QAM) modes. Finally, compared with the traditional algorithms based on Cyclic Prefix (CP) or general PN, only the lengths of part of PN were calculated, so the number of autocorrelations was reduced, and better performance could be guaranteed. The simulation results show that, at Bit Error Rate (BER) of 10 ^-4, the improved algorithm has about 5 dB and 1 dB gains while comparing with the algorithms based on CP and general PN, respectively. And compared with algorithm with general PN, when the lengths of inserted sequence and cyclic sequence were 420 and 165, the number of correlations per frame was reduced by 1186. The theoretical analysis and simulation results show that proposed algorithm can effectively reduce the computational complexity and cost of process, meanwhile improves the communication rate.

To deal with the discrete series protocol message without session information, a new Separate Protocol Message based Format Signature Construction (SPMbFSC) algorithm was proposed. First, separate protocol message was clustered, then the keywords of the protocol were extracted by improved frequent pattern mining algorithm. At last, the format signature was acquired by filtering and choosing the keywords. Simulation results show that SPMbFSC is quite accurate and reliable, the recognition rate of SPMbFSC for six protocols (DNS, FTP, HTTP, IMAP, POP3 and IMAP) achieves above 95% when using single message as identification unit, and the recognition rate achieves above 90% when using session as identification unit. SPMbFSC has better performance than Adaptive Application Signature (AdapSig) extraction algorithm under the same experimental conditions. Experimental results indicate that the proposed SPMbFSC does not depend on the integrity of session data, and it is more suitable for processing incomplete discrete seriesprotocol message due to the reception limitation.

With the deepening of Weibo research, credibility evaluation of Weibo users has become a research hotspot. Aiming at the problem of Weibo users' credibility evaluation, a user confidence analysis method based on association was proposed. Taking Sina Weibo as the research object, firstly, seven characteristics of the user from three aspects: user information, interactive information and behavior information were analyzed, and the user self-evaluation credibility was got by using Analytic Hierarchy Process (AHP). Then, by using the user self-evaluation as the base point, the user relationship network as the carrier, and the potential users' evaluation relationship among the users, was improved the PageRank algorithm, and the user credibility evaluation model called User-Rank was proposed. The proposed model was used to evaluate comprehensively credibility of users by other users in relational network. Experiments on large scale Weibo real data show that the proposed method can obtain good evaluation results of user credibility.

Feature representation and metric learning are fundamental problems in person re-identification. In the feature representation, the existing methods cannot describe the pedestrian well for massive variations in viewpoint. In order to solve this problem, the Color Name (CN) feature was combined with the color and texture features. To extract histograms for image features, the image was divided into zones and blocks. In the metric learning, the traditional kernel Local Fisher Discriminant Analysis (kLFDA) method mapped all query images into the same feature space, which disregards the importance of different regions of the query image. For this reason, the features were grouped by region based on the kLFDA, and the importance of different regions of the image was described by the method of Query-Adaptive Late Fusion (QALF). Experimental results on the VIPeR and iLIDS datasets show that the extracted features are superior to the original feature; meanwhile, the improved method of metric learning can effectively increase the accuracy of person re-identification.

In order to improve the transformation efficiency of tile-pyramid image, a 15-parameter projection transformation method was established by quartic polynomial based on the view model of digital earth. The influencing factors for selecting the size of tile image were discussed theoretically, and an optimization method to determine the size and depth of tile-pyramid was given. To test this algorithm, a basic digital earth environment BDE2 was constructed by adopting JOGL. The analysis and experimental results show that tile-pyramid in 10m pixel accuracy constructed by this algorithm only has 10 layers and less than 5×10-5 average error; meanwhile, the proposed algrithm has low complexity, close stitching, high definition and low distortion, and can effectively avoid stitch cracks and characteristics distortion after the image is transformed.

Loop program has a significant amount of execution time in digital signal processing software, temporary storage of loop code with instruction buffer can reduce the number of program memory access to improve the performance of processor. A loop instruction buffer was added in the instruction pipeline. It could store and dispatch instructions of loop program in the software pipelining manner. The instructions of loop program needed to be accessed from program memory only once but executed many times, so the number of memory access was reduced. During the loop instructions were dispatched from buffer, the program memory could be signaled to sleep to reduce the power consumption of processor. In the typical application program, the instruction pipeline can be idle above 90%, and the performance of processor is improved about 10%, the overhead of loop buffer is 9% of the instruction pipeline.

A low power branch encoding method was presented for decreasing the SoC bus power dissipation. This method's basic principle is: for the address bus, when the address bus is sequential, the address bus is frozen, and when the address bus is non-sequential, the window size is adjusted dynamically to apply the Bus-Invert (BI) method on the address bus. For the data bus, two threshold values are figured out for different data size respectively. If the Hamming distance locates between these two threshold values, the valid-data-channel switching dense area is found and inverted, otherwise applies the BI encoding. This method's encoding and decoding circuits are realized in the Advanced High Performance Bus (AHB) system. The experimental result demonstrates that compared with uncoded situation, this method decreases the address/data bus toggle rate by 51.2%/22.4%, and the system power is reduced by 28.9%. Compared with T0,BI and other encoding methods realized in the same system, the branch encoding is more superior in the toggle rate and power dissipation.

In order to reduce the test application time and guarantee high test data compression rate, a selection sequence of parallel folding counter was proposed. Selection test sequences were generated by recording group number and in-group number which represented folding index based on the analysis of parallel folding computing theory, so as to avoid generating useless and redundant test sequences. The experimental results on ISCAS benchmark circuits demonstrate the average test compression rate of the proposed scheme is 94.48%, and the average test application time is 15.31% of the similar scheme.

To solve the low extraction efficiency for extracting hot topics in huge amounts of micro-blog data, a new topics detection method based on user role orientation was proposed. Firstly, some noise data of parts of users were filtered out by user role orientation. Secondly, the feature weight was calculated by the Term Frequency-Inverse Document Frequency (TF-IDF) function combined with semantic similarity to reduce the error caused by semantic expression. Then, the improved Single-Pass clustering algorithm was used to extract the topics of micro-blog. Lastly, the heat evaluation of micro-blog topics was made according to the number of reposts and comments, thus the hot topics were found. The results show that the average missing rate and false detection rate respectively decrease by 12.09% and 2.37%, and further indicate the topic detection accuracy rate is effectively improved and the method is feasible.

Node deployment is a basic problem in sensor networks, which directly relates to the performance of the entire network. Most existing researches on sensor network node deployment are for the case of twodimensional planar and three dimensions space, but very few researches for threedimensional surface deployment scenario. This paper proposed an algorithm of optimal surface deployment in wireless sensor networks. First by mathematical or differential geometry method for threedimensional surface it constructed mathematical model, and then through the centroid of the threedimensional surface Voronoi subdivision partitions, an error function was proposed to evaluate the superiority of deployment method. Finally compared with other surface deployment methods, the performance of the proposed algorithm in this paper is superior.

Concerning the defects of multi-scale threshold on directional information using Contourlet transform,a new adaptive threshold denoising algorithm was proposed, which was based on average energy of neighboring window. According to the distribution of the coefficients energy, the Contourlet coefficients were divided up into three areas. The noise could be reduced obviously by adjusting the threshold of these areas with different variables. In contrast with the wavelet threshold, Contourlet threshold and multi-scale threshold using Contourlet transform, the experimental results demonstrate that the new algorithm has superiority in Peak Signal-to-Noise Ratio (PSNR) and visual effect, which can maintain effectively the edges of the image.

In order to improve the face detection test results under the influence of illumination change and complex background, an algorithm of 3D color clustering model based on direct least squares estimate was proposed during the preprocessing phrase. Firstly, three plane projection distributions of skin color were seen as fitting objects in CbCrCg space, and then smooth edge was got by median filter and Sobel operator, at last the best 3D color model was got through direct least squares. In experiment, the public face library and face image got by outdoor shooting were seen as objects, and the experimental results show that, this algorithm has better segmentation effects than traditional color preprocessing algorithm, and it has improved the detection rate more effectively.

Since the present network security assessment methods cannot evaluate vulnerability relevance effectively, a vulnerability threat assessment method based on relevance was presented. Firstly, an attack graph must be created as the source data. Secondly, by taking both pre-nodes and post-nodes diversity into consideration, integrating the methods of Forward In (FI) and Backward Out (BO), the authors calculated the probability of vulnerability being used on multiple attack routes through optimizing calculation formulas originating from Bayesian network, then the weighted average method was utilized to evaluate the risk of certain vulnerability on a particular host, and finally the quantitative results were achieved. The experimental results show that this method can clearly and effectively describe the security features of systems.

In trust computing, recommendation trust has very strong subjectivity and some aggressive behaviors like deception and slander. Those factors will conceal the authenticity of behaviors of user who is recommended and threat the system security. To address the problem, this paper proposed a weighted trust computing method based on user’s behaviors. The time attribute of feedback information was identified by using time attenuation. And the trustworthiness of users was computed based on directness trust and recommendation trust with different weights. Also, this paper introduced Feedback credibility to evaluate the authenticity of recommendation trust. Simulation experiments show that this method has better adaptability to the dynamics of trust. It can reduce effectively the impact of malicious recommended trust, and compute accurately the trust of users according to user’s behaviors, which provide reliable information to correctly make security decision for the system.

Concerning the problem of low traversal efficiency when searching the FP-tree for conditional pattern bases, a new no-header-table FP-Growth algorithm was proposed. The algorithm employed a recursively backtracking way to search for conditional pattern bases, avoiding traversing the same FP-tree path multiple times. Compared with FP-Growth algorithm in terms of theoretical analysis and actual mining performance, this algorithm greatly reduced the searching cost and improved the mining efficiency of frequent patterns by 2-5 times. Finally, the algorithm was used to mine association rules in telecommunication network alarms. The high mining speed, with the coverage of 83.3% against correct rules, shows that it is superior to FP-Growth both in time and space performance.