The existing image captioning methods only focus on the grid spatial location features without enough grid feature interaction and full use of image global features. To generate higher-quality image captions， a global image captioning method based on Graph ATtention network （GAT） was proposed. Firstly， a multi-layer Convolutional Neural Network （CNN） was utilized for visual encoding， extracting the grid features and entire image features of the given image and building a grid feature interaction graph. Then， by using GAT， the feature extraction problem was transformed into a node classification problem， including a global node and many local nodes， and the global and local features were able to be fully utilized after updating the optimization. Finally， through the Transformer-based decoding module， the improved visual features were adopted to realize image captioning. Experimental results on the Microsoft COCO dataset demonstrated that the proposed method effectively captured the global and local features of the image， achieving 133.1% in CIDEr （Consensus-based Image Description Evaluation） metric. It can be seen that the proposed image captioning method is effective in improving the accuracy of image captioning， thus allowing processing tasks such as classification， retrieval， and analysis of images by words.

To solve the problem that the direct use of high-dimensional， high-frequency， noise-containing real-world data to perform data processing leads to unreliable estimators， a data uncertainty quantification method based on Generative Adversarial Network （GAN） was proposed. Firstly， the original data distribution was reconstructed by GAN to construct a mapping distribution from the noise space to the space of the original data. Secondly， the samples were extracted by Markov Chain Monte Carlo （MCMC） method to obtain new samples based on the original data distribution. Thirdly， confidence intervals for the uncertainty of the samples were defined based on the specified functions. Finally， the confidence intervals were used to estimate the uncertainty of the original data， and within the data the confidence intervals was selected as the data used by the estimator. Experimental results show that 50% fewer samples are required to train the estimator to reach the upper limit by using the data within the confidence intervals compared to the samples required by using the original data. At the same time， compared to the original data， the data within the confidence intervals requires 30% fewer samples on average to achieve the same test accuracy.

With the continuous increment of data and system size， network communication becomes a performance bottleneck of key-value storage systems. Meanwhile， Remote Direct Memory Access （RDMA） technique can support high bandwidth， low latency data transmission， which provides a new idea for designing key-value storage systems. Based on RDMA technique in the high performance network， a key-value storage system named Chequer with high performance and low CPU overhead was designed and implemented. By combining the characteristics of RDMA primitives， the basic operation workflow of key-value storage system was redesigned. And a linear probing based shared hash table was designed to reduce the number of client reading rounds by solving the problem of client cache invalidation as well as increasing the hash hit rate， which can further improve the performance of the system. The Chequer system was implemented on the small-scale cluster， and its performance was demonstrated by experiments.

In order to improve the conventional situation that network video does not play smoothly during live or on-demand and find storage strategy of mass video data, this paper presented an overall design scheme of a real-time live video recording and storage system. The open source streaming media server Red5 and the rich Internet application technology Flex were utilized to achieve live video streaming and recording. The recorded video data would be stored in the open source NoSQL database MongoDB. The experimental results illustrate that the platform can meet requirements of multi-user access and data storage.〖JP〗

Considering that contact duration has significant influence on whether packet can be transmitted successfully or not, the authors proposed a Probabilistic Routing Protocol using History of Encounters and Transitivity based on Contact Duration (PRoPHET-CD), which combined contact duration with encounter frequency to estimate delivery probability. This protocol could improve the delivery probability significantly and reduce the interruption of packet transmission. The simulation results show that the protocol of PRoPHET-CD can significantly enhance the message delivery probability and reduce the overhead ratio.

Active queue management is a focus of current research. Random Early Detection (RED) is one kind of classical queue management algorithms. Linear RED is simple and easy to calculate; however, when average queue size is near to the minimum and maximum threshold, the loss rate is unreasonable. After verifying the nonlinear character between average queue size and packet loss rate, an improved RED algorithm named JRED was presented. The simulation on NS2 shows that the average throughput is improved, and the packet loss rate is decreased. With the JRED algorithm, the stableness and reliability of network are enhanced.

A definition of universally composable security of identity-based signature schemes is proposed in the universally composable security framework. The equivalence of the universally composable security and the traditional security of identity-based signature schemes is proved. This result shows that an identity-based signature scheme can be used as a primitive block to design more complicated cryptographic protocols.

Petri net is an effective mathematical tool for formally analyzing and verifying protocols. On the basis of the analysis of Low-Energy Adaptive Clustering Hierarchy (LEACH) protocol, a Generalized Stochastic Petri Nets (GSPN) model of performance for LEACH protocol was proposed. With the help of SPNP tools for analyzing the performance model of protocol, the validity of the model was verified. Finally, the significance of the performance model to the protocol's low-energy improvement was discussed.

Based on Petri net, a new consistent bothbranch varyingweight fuzzy Petri attack net (CBBVFPN) was defined using both-branch fuzzy set theory, and a formal reasoning algorithm for the CBBVFPN model was presented according to the characteristics of CBBVFPN model. The model makes up for some common limitations in previous attack models based on Petri net, and two basic structures about "AND" and "OR" of Petri net were expanded. In reasoning process, in order to avoid continuing to reason by taking a proposition with very low truth degree as a precondition, and then a post proposition with lower truth degree was got, a rule evolvement and varying-weight method was presented. The reasoning algorithm described positive and negative propositions in a unified way, and added the controlling conditions in the algorithm to ensure the consistency of reasoning processes and reasoning results. Finally, a Botnet attack was used to verify the reasoning algorithm.

An efficient multisignature scheme was proposed, whose security is tightly related to the hardness of the Dicisional DiffieHellman problem (DDHP) in the random oracle model. Under the Proof-Of-Possession (POP) key registration assumption, the scheme was proved unforgeable against the adaptive chosen message and signing groups attacks. Both the size of multi-signature and the computation cost of verification do not increase with the increase of the signers. Compared with the Computational Diffie-Hellman Problem (CDHP)-based scheme with tight security reduction proposed by Z C WANG, the scheme has the merits of fewer communication rounds and less computation cost. Thus it can be used in many electronic transactions.

The time synchronization mechanism is one of the key problems on Agent-based distributed simulation research. The time synchronization mechanism of Agent-based distributed simulation environment (ADSimE) — conservative mechanism and optimistic mechanism were introduced. That is, the conservative mechanism was used within the node of the platform，while optimistic mechanism was used between the nodes. This kind of time synchronization mechanism could promote the performance of the time advancing in the simulation platform. The principles of the two mechanisms were also introduced. Finally, the key process of the design and running for conservative simulation Agent and optimistic simulation Agent were described in detail.

In order to overcome the defect that the conventional color histogram retrieval method is prone to lose the spatial information of colors, an image retrieval method based on the largest color-connected regions was presented, and its rim roughness to capture the regional traits and reduce the absolute dependence on color. Then the similarity of the images is defined as the combination of the similarity of an integer and the similarity of the blocks to reflect the recognition process by people. The experimental result on real-world image collections indicates that the proposed approach significantly improves image retrieval accuracy.

Establishing a mapping relationship between image visual features and semantics can be used to reduce "semantic gap". A novel mapping method for image semantics and visual features was presented. In this method, image semantic information could be captured by adding users relevance feedback, and then a decision table of visual features and semantics was constructed. Knowledge reduction of rough set theory was used to reduce the redundant visual features according to semantics, by doing that a mapping relationship between image visual features and its semantics was established. The experimental results indicate that amount of visual features irrelevant to image semantics can be reduced greatly, the complexity and cost of semantic classification are reduced and the accuracy of classification is better.

The security weakness of structured multi-signature scheme with signers' intentions was pointed out. It is that the signers' intentions could be modified by an inner or outer attacker, thus the security of this type of signature is tampered. By using a more secure signing algorithm to sign the signers' intentions, the problem was solved. The improved scheme not only has all the necessary security properties of this type of signature, but also has improvement in efficiency - shorter signature size, more efficient in communication and signing computation. The shortage of the new scheme is some decrease in the efficiency of verification.

Stereo pair acquisition of a scene is the key to binocular stereo imaging. This paper presented a stereo pair creation method when 3D models were constructed. Using camera objects in 3DS MAX, the method started from a coordinate transformation of objects in the scene based on principle of binocular stereo vision. Then the method carried out the perspective transformation to create left image and right image respectively. The results of the experiment indicate the position of the two target cameras and the 3D model, together with the length of the base line is the key factor that affects the stereo effect. Changing the position of the target cameras and the 3D model may result in positive disparity or negative disparity stereo pairs. When the aspect ratio of AB to CO equals 0.05, the stereo effect of the stereo pairs created is better.