The existing deep learning-based methods for source code vulnerability detection often suffer from severe loss of syntax and semantics in target code， and neural network models allocating weights to the graph nodes （edges） in target code unreasonably. To address these issues， a method named VulATGCN for detecting source code vulnerabilities was proposed on the basis of Code Property Graph （CPG） and Adaptive Transformer-Graph Convolutional Network （AT-GCN）. In the method， CPG was used to represent source code， CodeBERT was combined for node vectorization， and graph centrality analysis was employed to extract deep structural features， thereby capturing the code’s syntax and semantic information in multi-dimensional way. After that， AT-GCN model was designed by integrating strengths of Transformer-based self-attention mechanism， which excels at capturing long-range dependencies， and Graph Convolutional Network （GCN）， which is proficient at capturing local features， thereby realizing fusion learning and precise extraction of features from regions with different importance. Experimental results on real vulnerability datasets Big-Vul and SARD show that the proposed method VulATGCN achieves an average F1 score of 82.9%， which is 10.4% to 132.9% higher than deep learning-based vulnerability detection methods such as VulSniper， VulMPFF， and MGVD， with an average increase of approximately 52.9%.

The defects such as sidewall foreign matter, crown foreign body, air bubble, crown split and sidewall root opening that appear in the process of tire production will affect the use of tires after leaving factory, so it is necessary to carry out nondestructive testing on each tire before leaving the factory. In order to achieve automatic detection of tire defects in industry, an automatic tire defect detection method based on improved Faster Region-Convolutional Neural Network (Faster R-CNN) was proposed. Firstly, at the preprocessing stage, the gray level of tire image was stretched by the histogram equalization method to enhance the contrast of the dataset, resulting in a significant difference between gray values of the image target and the background. Secondly, to improve the accuracy of position detection and identification of tire defects, the Faster R-CNN structure was improved. That is the convolution features of the third layer and the convolution features of the fifth layer in ZF (Zeiler and Fergus) convolutional neural network were combined together and output as the input of the regional proposal network layer. Thirdly, the Online Hard Example Mining (OHEM) algorithm was introduced after the RoI (Region-of-Interesting) pooling layer to further improve the accuracy of defect detection. Experimental results show that the tire X-ray image defects can be classified and located accurately by the improved Faster R-CNN defect detection method with average test recognition of 95.7%. In addition, new detection models can be obtained by fine-tuning the network to detect other types of defects..

Concerning the problem that the calculations of the existing regenerating code schemes is based on GF( q), and it has high computational complexity and low efficiency, a regenerating code construction method based on sparse random matrix over GF(2) and product matrix framework was proposed. Firstly, file data was arranged in a matrix and the row XOR operation was performed according to encoding matrix. Secondly, local data was encoded by helper nodes according to failed node's encoding vector and sent to repair node. Finally, the failed node's data was decoded by repair node according to received data. The experimental results show that the repair bandwidth of the proposed method is only one-tenth of traditional erasure code at most, and the encoding rate increases by 70% and decoding recovery rate increases by 50% compared with regenerating code based on conventional Vandermonde matrix, which facilitates the application of regenerating code in massive storage system.

Focusing on inefficient data security and access control in the existed cloud storage, which results in sensitive information to be stolen, combined with the existed Ciphertext-Policy Attribute-Based Encryption (CP-ABE) and data partition,an efficient data privacy protection model based on the hybrid cloud was proposed. First of all, according to the data sensitive degree, the data were divided into data blocks based on different sensitivity levels, and then data blocks were stored on different cloud platforms. According to the security level of the data, data were encrypted by using the different intensity encryption technologies. At the same time, the scheme of "first match after decryption" was adopted in the decryption stage and the algorithm was optimized. Finally, user decrypted ciphertext by the multiplication. Compared with the single node algorithm, for encrypting 1 Gb data, the efficiency of symmetric encryption algorithm more than doubled in the public clouds. The experimental results show that the proposed scheme can protect the privacy data of cloud storage user, reduces the system cost and improves the system flexibility.

Concerning the problem that the linear eigentransformation method cannot capture the statistical properties of the nonlinear facial image, a Data-driven Local Eigentransformation (DLE) method for face hallucination was proposed. Firstly, some samples most similar to the input image patch were searched. Secondly, a patch-based eigentransformation method was used for modeling the relationship between the Low-Resolution (LR) and High-Resolution (HR) training samples. Finally, a post-processing approach refined the hallucinated results. The experimental results show the proposed method has better visual performance as well as 1.81dB promotion over method of locality-constrained representation in objective evaluation criterion for face image especially with noise. This method can effectively hallucinate surveillant facial images.

A new adaptive neighborhood Multi-Objective Grid Task Scheduling Algorithm (ANMO-GTSA) was proposed in this paper for the multi-objective job scheduling collaborative optimization problem in grid computing. In the ANMO-GTSA, an adaptive neighborhood method was applied to find the non-inferior set of solutions and maintain the diversity of the multi-objective job scheduling population. The experimental results indicate that the algorithm proposed in this paper can not only balance the multi-objective job scheduling, but also improve the resource utilization and efficiency of task execution. Moreover, the proposed algorithm can achieve better performance on time-dimension and cost-dimension than the traditional Min-min and Max-min algorithms.

In order to assess the risk of network security more reasonably, a cloud-model based method for network risk assessment was proposed. It took advantage of cloud model featuring perfect combination of randomness and fuzziness. Firstly, standard clouds were constructed by sampling normal system status. When making risk assessments, the current risk state was sampled to calculate the cloud characteristics, then the cloud similarity algorithm based on the distance measurement of cloud droplets was used to calculate the similarity between them, and the biggest similarity was the final output. Finally, Kddcup99 data set was used to do simulated attack and performance sampling test. The experimental results show that the proposed method retains the maximum uncertainty of network intrusion assessment and improves the credibility of the results.

In order to improve the speed of global optimization algorithm, a global optimization algorithm of intelligence control system was presented. The feedback idea of closed loop control system was applied in this algorithm that made the value of the object function gradually close to the input in the iterative process until reaching the global optimization. The key of the algorithm lies in the design of control strategy and the initial setting of parameters. In order to reduce the difficulty of initial setting of parameters and ensure the precision of the algorithm, the filled function was used to improve the global optimization algorithm of intelligence control system. Verified by twelve standard test functions, the improved algorithm is faster than filled function method, and is more accurate than the global optimization algorithm of intelligence control system.

VIrtual Scanning Algorithm（VISA) is unable to fully take advantage of the number of nodes, in order to prolong the network lifetime, it must be built on the basis of dense nodes deployment which makes the time for finding average target increase. Therefore, based on low dutycycle Wireless Sensor Network (WSN) by combining the ideology of virtual scan wave, the Multiple VIrtual Scan Algorithm (M-VISA) was proposed for road surveillance. This algorithm adopted the way of fixing points, deploying the same location with multinodes to let the nodes to be worked in order and in batches, hence, the network lifetime could be greatly extended. Simulation result demonstrates that M-VISA can prolong network lifetime by 180% when compared with VISA, improving the network performance effectively.

To mine accurately process patterns and improve anti-noise ability, a new method of process pattern mining based on probability statistics of adjacent event was proposed to generate automatically process patterns mining based on rules through only once traversing logs and simply calculating matrix. Comparing with α-algorithm and heuristic algorithm, the results from analyzing mining performance and verifying experiment show that it can mine pattern structures of sequence, choice, paralleling, recursion and short loop, and has advantages of low complexity and good anti-noise ability.

Automatic analysis system of Chinese character structure applies many mature techniques in the fields such as computer vision, image understanding to analyze and recognize the Chinese character. GB2312-80 Chinese characters set was chosen to generate subcomponent images. Then the feature description of Chinese characters component structures is generated, and valid ratio is over 90%.