Search

Deep neural network (DNN) is applied widely in many applications and achieves state-of-the-art performance. However, DNN lacks transparency and interpretability for users in structure. Attackers can use this feature to embed trojan horses in the DNN structure, such as inserting a backdoor into the DNN, so that DNN can learn both the normal main task and additional malicious tasks at the same time. Besides, DNN relies on data set for training. Attackers can tamper with training data to interfere with DNN training process, such as attaching a trigger on input data. Because of defects in DNN structure and data, the backdoor attack can be a serious threat to the security of DNN. The DNN attacked by backdoor performs well on benign inputs while it outputs an attacker-specified label on trigger attached inputs. Backdoor attack can be conducted in almost every stage of the machine learning pipeline. Although there are a few researches in the backdoor attack on image classification, a systematic review is still rare in this field. This paper is a comprehensive review of backdoor attacks. According to whether attackers have access to the training data, we divide various backdoor attacks into two types: poisoning-based attacks and non-poisoning-based attacks. We go through the details of each work in the timeline, discussing its contribution and deficiencies. We propose a detailed mathematical backdoor model to summary all kinds of backdoor attacks. In the end, we provide some insights about future studies.

Object detection is one of the essential tasks of computer vision. Object detectors based on the deep neural network have been used more and more widely in safe-sensitive applications, like face recognition, video surveillance, autonomous driving, and other tasks. It has been proved that object detectors are vulnerable to adversarial attacks. We propose a novel black-box attack method, which can successfully attack regression-based and region-based object detectors. We introduce methods to reduce search dimensions, reduce the dimension of optimization problems and reduce the number of queries by using the Covariance matrix adaptation Evolution strategy (CMA-ES) as the primary method to generate adversarial examples. Our method only adds adversarial perturbations in the object box to achieve a precise attack. Our proposed attack can hide the specified object with an attack success rate of 86% and an average number of queries of 5, 124, and hide all objects with a success rate of 74% and an average number of queries of 6, 154. Our work illustrates the effectiveness of the CMA-ES method to generate adversarial examples and proves the vulnerability of the object detectors against the adversarial attacks.

In terms of timeliness and reliability, the performance requirements of the control and protection system of the Zhangbei 500kV hybrid high-voltage DC circuit breaker is put forward. To meet these performance requirements, the control and protection system suitable for 500kV hybrid high voltage DC circuit breakers is designed. The topology and working principle of Zhangbei hybrid highvoltage DC circuit breaker is analyzed. The basic circuit breaker control strategy and body protection configuration are studied. The hardware architecture and software flow of the control and protection system are designed. Furthermore, a series of designs such as high-performance and anti-electromagnetic interference of the board have been carried out. The redundancy design of the system is carried out from four aspects: control system, protection system, monitoring system and power supply. Finally, combined with the Zhangbei flexible DC power grid test demonstration project, the hybrid high-voltage DC circuit breaker control and protection system is verified by the current breaking test and the reclosing test. The test results show that the control cycle is $25\mu\mathrm{s}$ and a protection action delay is less than $350\mu\mathrm{s}$ . The system can reliably complete the breaking of the circuit breaker within 2.7ms, the impact of fault current on converter valves and other equipment and systems is reduced. The performance requirements of 500kV hybrid high-voltage DC circuit breakers in Zhangbei project are met.

The significant momentum of computer-bridge brings a strong demand for quality validation and assurance for computer-bridge software. Current research work seldom discusses corresponding software testing questions and validation approaches with clear quality requirements and methods. This paper focuses on a platform based on anonymous pipes that can load different game engines for multi-program battles or man-program battles, including the essential problems in the platform, features, and process. Moreover, it presents a test process for a certain computer-software. Finally, it discusses the challenges, issues, and needs in computer-software testing.

Smart grid is the development direction of the future power grid. A reasonable evaluation index system of smart grid can accurately evaluate the overall development and construction level of smart grid, and can measure the difference of the smart level of regional power grid. On the basis of studying the factors of power grid security, quality, efficiency and sustainability, this paper classifies and layers the multi-evaluation indexes in the smart grid evaluation index system based on the analytic hierarchy process (AHP), then gives the evaluation methods of each index, and finally obtains the overall evaluation results. Finally, a simulation example is given to illustrate the effectiveness of the proposed method.

Most of the existing Matcha blending is done manually, which is inefficient and leads to uneven tea products. To solve this problem, this paper proposes an intelligent blending method based on improved PID control algorithm. In this paper, considering the various factors that affect the quality of tea, without affecting the quality of the product, taking the lowest cost of the enterprise as the constraint goal, the basic combination model is constructed by using PID control algorithm. Since the selection of PID parameters will affect the output of the model, this paper introduces particle swarm optimization algorithm to optimize the PID controller parameters globally, so that the PID controller can achieve the ideal control index, and finally determine the optimal combination of Matcha. The experimental results show that the method proposed in this paper can effectively realize the intelligent blending of Matcha, and maximize the cost of tea factory, effectively improve the income of Matcha products, which is conducive to the further development of tea companies.

In the process of the continuous reform of modern social economic system and the continuous development of science and technology, the distribution of the Internet of things reflects good application value. However, the Internet of things has many distributed nodes, data center, high computing cost and poor supervision. In the process of privacy protection, there are still some objective problems. Therefore, the use of blockchain technology in the privacy protection of the Internet of things privacy data block, structure decentralization, information hierarchy and other aspects, which can establish a more perfect Internet of things privacy protection model and more rich management mechanism, making the prevention of privacy leakage, data tampering and other aspects of privacy protection work has been further developed. Therefore, this paper takes the implementation mode of blockchain smart contract in smart home privacy protection as an application example to analyze the application value of blockchain in the privacy protection of Internet of things.

Game robot is a challenge in the field of artificial intelligence. First of all, the current status of robot and computer game were described. After the techniques of computer games and the characteristics of intelligent robots were analyzed, a concept of intelligent robot was proposed, which can be used in computer games. Then the design and implementation methods were given. In the end, the importance of the kind of intelligent robots was discussed and several application prospects were provided in the future.

Surakarta chess is researched with the computer game technology, which is a kind of computer chess game in the complete information game. In order to increasing the human-computer interaction of the computer games system, the representing method of chessboard with 3D form is studied by use of unity 3D game engine technology. The Surakarta game system is designed using VC++ language. In this system, the Monte Carlo Tree Search(MCTS) algorithm and Minimax Algorithm are used together. This system has improve the visual effect, and combining two searching algorithm is feasibility.

A platform of online bridge card game is designed for computer games in this paper. To implement a real-time platform for sharing information, we use TCP/IP network transport protocol to realize it, and data for transmission is very few. MFC development framework is used for programming. By using the platform, it is easy to develop other game platform and AI programs in a short time. So the research of game platform has some reference significance to the development of the computer games.

Computer Game is one of the important and challenging research directions in the field of Artificial Intelligence (AI). This paper introduces the composition of the computer game system and the rules of Checkers (100) that is a complex two-player chess game. The characteristics of the evaluation of the Checkers are described in this paper. Four main features and value matrix of the pieces are also mainly elaborated. According to the importance of the different positions of the Checkers, an evaluation function with value matrix of the pieces was proposed. The experimental results show that the proposed evaluation function is feasible and effective.

This paper analyzes the traditional game system, pointing out its shortcoming and putting forward an idea of designing a Distributed System (which is referred to as DS) Computer Game Platform. At first, the paper puts forward the structure of DS computer game platform and introduces several functions. After that, the paper proves the practicability of DS computer game system in several different aspects such as the technological principles, running processes and architectural design. Finally, comparing with the traditional form of playing computer game, the paper proves that the DS Computer Game Platform can be more powerful and efficient.

This paper simply introduces the game of Amazon and traditional alpha-beta pruning algorithm, mainly introduces the using of UCT algorithm in Amazon, and using MC evaluation to replace traditional evaluation function. We also establish a move table to speed up the search efficiency. This paper compares UCT algorithm with alpha-beta pruning algorithm, and analyses the different performance of each algorithm in different stages of games.

Aiming at multi-focus color image fusion, An image fusion algorithm based on the Canny edge detection is proposed. Firstly, the color image is converted into grayscale image, then, the Canny edge detection algorithm is used to detect the edge of grayscale image to obtain the corresponding edge image. Secondly, the area of the two edge images is compared and the isolated point is removed. The stripes on the non-focus area of the edge image are eliminated. Then, the black areas of the two images are compared, and the simple decision map is gotten. Finally, the decision map is processed by processing the edge. The value of the edge point is re-determined, and the final decision map is obtained. The multi-focus fusion image is obtained according to the decision map. The experimental results show that, the proposed algorithm is superior to traditional fusion algorithms, in terms of objective criteria and visual appearance.

Phantom Go as an important branch of computer games is emerging. Monte Carlo algorithm is widely used in this kind of incomplete information games currently. However, there is a tradeoff between run time and its effectiveness. In this paper, based on the characteristics of Phantom Go, a novel algorithm is put forward, which merges Open Multi-Processing and Compute Unified Device Architecture parallel computing methods into the Monte Carlo algorithm. It improves the ability of Phantom Go to some extent.

This paper mainly introduces the structural model used for achieving the two-person game platform on computer games. To analyze the realistic logical processes of computer games, the referee-player role model is induced, computer game platform makes it true and three focal technologies is proposed, that is, move event driven in game process, target player orientation in referee instructions and interaction between equivalent cores.

This paper comes up with the framework design and detailed technical analysis on the intelligent decision-making system of the computer game of checkers, and the prospect towards the future technology in research. The system includes the following four parts: the design and expression of the board, the achievement of the basic move, the establishment of the valuation system, and the construction of the search engine. After the experiment of a match between man and machine, and also that between machine and machine, the system can implement the function of intelligent and optimal decision-making. This system contributes a lot to the intelligent decision-making theory and the application of its algorithms on the computer games technology.

An Effective low-light-television (LLTV) and forward-looking-infrared (FLIR) aerial images fusion method based on dual-tree complex wavelet transform (DT-CWT) is presented for flight navigation. The deficiencies of traditional discrete wavelet transform (DWT) based algorithms in characteristics representation are very obvious. The DT-CWT is suitable to image fusion processing because of its shift invariance and directional selectivity. It is preferable to traditional DWT. The simple and effective fusion rules are also putted forward. The high frequency parts use absolute value as measurement, and the low frequency parts use average after DT-CWT of source images. Comparing with the other seven methods, the experimental results and the evaluating statistics demonstrate that the proposed fusion algorithm is effective.

Due to the large scale of game tree, AI computer game programs run slowly in the debug mode of Microsoft vc++6.0 IDE and is very hard to debug the core algorithms. This paper argues that, based on Zobrist hashing and other technologies, we could do debugging works in release mode rather then in debug mode to greatly speed up the program development. An example is given about how to debug a Connect6 computer game named Consau with this method.