Botnets have become one of the main threats to cyberspace security. Although they can be detected by techniques such as reverse engineering, botnets using covert technologies such as fast-flux can successfully bypass existing security detection and continue to survive. The existing fast-flux botnet detection methods are mainly divided into active and passive, the former will cause a large network load, and the latter has the problem of cumbersome feature value extraction. In order to effectively detect fast-flux botnets and alleviate the problems in traditional detection methods, a fast-flux botnet detection method based on spatiotemporal features of network traffic is proposed, combined with convolutional neural networks and recurrent neural network models, the fast-flux botnet is detected from both spatial and temporal dimensions. Experiments performed on the CTU-13 and ISOT public data sets show that compared with other methods, the accuracy rate of the proposed method is 98.3%, the recall rate is 96.7%, and the accuracy is 97.5%.

In order to solve the problem of the long decoding delay for the Spatially-Coupled Low-Density Parity-Check (SC-LDPC) code with long code length, a Layered Sliding Window Decoding (LSWD) algorithm is proposed. By exploring the quasi-cyclic characteristics of the SC-LDPC sub-codeblock and the hierarchical structure of the check matrix in the sliding window, the part of check matrix in the sliding window is layered to optimize the message transfer between two neighbor layers, with the aim of accelerating the convergence of the iterative procedure and reducing the number of decoding iterations. Simulation and analysis results show that the number of iterations in the proposed LSWD algorithm is less than that in the SWD, under the same Signal-to-Noise Ratio (SNR) and the bit error ratio. In the high SNR region, especially, the number of iterations in the proposed LSWD is about half of that in the SWD, hence the global decoding delay of the former is effectively shorten. In addition, the decoding performance of the LSWD algorithm is better than the SWD algorithm under the same number of decoding iterations, and the overall computational complexity is slightly increased.

In view of the problem of high complexity for non-sinusoidal time domain modulation algorithms based on Prolate Spheroidal Wave Functions (PSWFs), spatial mapping is introduced to analyze the complete orthogonality and derive the minimum number of sampling points of PSWFs in the frequency domain. On this basis, the complex domain mapping and FFT/IFFT signal processing framework are introduced, and the PSWFs frequency domain modulation and demodulation method are proposed. The proposed method extends PSWFs signal processing from time domain to frequency domain, providing a possibility for exploring and studying the application of PSWFs signal to 5G, beyond 5G which use frequency domain signal processing. Theory and numerical analysis show that, compared with the time domain modulation, the proposed method can reduces the complexity of the algorithm from O(2Qg²) to O(g²+glog₂g) without changing the system spectral efficiency, system error performance, modulation signal energy aggregation, and peak-to-average power ratio.

In order to reduce the computation complexity and storage capacity of the Kernel Affine Projection P-norm (KAPP) algorithm, and improve the convergence rate and steady-state performance of the algorithm when the input signal is strongly correlated, a Kernel Normalization Decorrelated Affine Projection P-norm algorithm based on Gaussian Kernel Explicit Mapping (KNDAPP-GKEM) is proposed. The correlation of the input signal is eliminated in advance by the normalized correlation method. The explicit kernel function is approximated by Gaussian kernel explicit mapping method, which eliminates the dependence on historical data and solves the problem that the computation and storage capacity of the KAPP algorithm are too high due to the continuous growth of structure. The simulation results of nonlinear system identification under α-stable distribution noise environment show that when the training data scale is large, the KNDAPP-GKEM algorithm still maintains a fast convergence rate and the low identification mean square error of nonlinear system. Moreover, its training time is linearly and slowly increased, which is more conducive to the practical application of nonlinear system identification.

The major drawback of Short Reference Differential Chaos Shift Keying (SR-DCSK) system is the low data transmission rate. To solve the problem, a Short Reference Multi-User Differential Chaos Shift Keying (SR-MUDCSK) communication scheme is proposed. The orthogonality of Walsh code is used to transmit the information of multiple users, which effectively improves the data transmission rate. The theoretical Bit Error Rate (BER) performance is analyzed, and experimental simulation is carried out under AWGN and multipath Rayleigh fading channel environment, respectively. The results show that the system has a significant improvement in transmission rate, and the energy efficiency is also significantly improved. Therefore, it is of great value in application.

In view of the lack of deployment and management of slicing in vehicular network, a slice coordination agent of vehicular network slicing structure is designed. Firstly, based on the K-means++ clustering algorithm, the vehicle network communication services are clustered according to the similarity and then mapped into different slices. Secondly, considering the imbalance of radio resource utilization caused by the space-time characteristic among application scenarios, a shared proportional fairness scheme is proposed to utilize radio resources efficiently and differently. Finally, in order to ensure the requirements of slicing service, linear programming obstacle method is used to solve the optimal slice weight distribution to maximize the slice load variation tolerance. Simulation results show that the shared proportional fairness scheme has smaller average Bit Transmission Delay (BTD) than the static slicing scheme, and the optimal slice weight distribution can be obtained under different user load distribution scenarios. The BTD gain achieves 1.4038 in the uniform user load scenario with 30 users per slice.

Considering the problem of agents’ network selection for Human-to-Human(H2H) and Machine-to-Machine (M2M) traffic in heterogeneous wireless networks, an agents’ network selection scheme based on the characteristic of traffic is designed. Game theory is adopted to solve the problem of network selection to satisfy difference in traffic’s Quality of Service (QoS) requirements. The existence and feasibility of the Nash Equilibrium (NE) of the proposed game are also analyzed. Then, a Distributed Network-Channe Selection Algorithm based on Learning Automata (DNCSALA) is presented to obtain the NE of the proposed game. In simulations, the proposed algorithm can achieve a near optimal performance compared to the exhaustive search, satisfy the QoS requirements of different types of traffic, and improves the efficiency of network resources.

For the problems of low delay, low power requirement and access congestion caused by computational unloading of mass devices, a Joint Offloading Decision and Resource Allocation Algorithm (JODRAA) is proposed based on cloud-fog hybrid network architecture. Firstly, the algorithm considers the combination of cloud and fog computing, and establishes a resource optimization model to minimize system energy consumption and resource cost with maximum delay as constraint. Secondly, the original problem is transformed into a standard Quadratically Constrained Quadratic Program (QCQP) problem, and a low-complexity joint unloading decision-making and computational resource allocation algorithm is designed. Furthermore, considering the access congestion problem caused by massive computing of unloading devices, an estimation model of the overflow probability of unloading user access request queue is established, and an on-line measurement based time-frequency resource allocation algorithm for fog nodes is proposed. Finally, the iterative bandwidth and power allocation strategy is obtained by using fractional programming theory and Lagrange dual decomposition method. The simulation results show that the proposed algorithm can minimize the system energy consumption and resource cost on the premise of time delay.

The application of Unmanned Air Vehicles (UAV)-enabled Cognitive Radio (CR) is widely used due to the convenience and high mobility of the UAV. In the UAV-based Cognitive Radio Network (CRN), the throughput optimization scheme in single radian is firstly investigated, in which the sensing radian is optimized to maximize the average throughput of UAV. Then, a multi-radian throughput optimization scheme based on Cooperative Spectrum Sensing (CSS) is studied to improve the sensing performance under the non-ideal channel, and the throughput of the UAV is maximized by utilizing an Alternative Iterative Optimization (AIO) algorithm. The simulation results show that the proposed scheme has better performance on improving the throughput of the UAV and ensuring the Quality-of-Service (QoS) of the Primary User (PU) when the channel fading is serious.

The generative adversarial network receives extensive attention in the study of computer vision such as image fusion and image super-resolution, due to its strong ability of generating high quality images. At present, the remote sensing image fusion method based on generative adversarial network only learns the mapping between the images, and lacks the unique Pan-sharpening domain knowledge. This paper proposes a remote sensing image fusion method based on optimized generative adversarial network with the integration of the spatial structure information of panchromatic image. The proposed algorithm extracts the spatial structure information of the panchromatic image by the gradient operator. The extracted feature would be added to both the discriminator and the generator which uses a multi-stream fusion architecture. The corresponding optimization objective and fusion rules are then designed to improve the quality of the fused image. Experiments on images acquired by WorldView-3 satellites demonstrate that the proposed method can generate high quality fused images, which is better than the most of advanced remote sensing image fusion methods in both subjective visual and objective evaluation indicators.

In order to solve the problems of lower precision of target location in short-term occlusion and inaccurate of scale estimation of target in rotation by Spatial-Temporal Regularized Correlation Filters (STRCF), an object tracking algorithm with channel reliability and target response adaptation is proposed in this paper. In this algorithm, target response regularization is added to train target model. By updating the ideal target response function in the process of solving model, the target can be tracked again after being occluded for a short time. The reliability of each feature channel is evaluated by coefficient of channel reliability, which can improves the model's expression of the target. Scale filters can be trained in log-polar coordinates to improve the accuracy of scale estimation when target is rotating. The experimental results show that the proposed algorithm reduces 28.54 pixels in the average center position error and improves the average overlap rate by 22.8% compared with STRCF.

A detection algorithm based on spatio-temporal context information is proposed to reduce the influence of non-uniform illumination and random jitter on the accuracy of target hole detection. The light equalization is carried out by using the spatial context information of target and its neighborhood, and the temporal motion context information between chest bitmap sequences is extracted for dithering correction. In order to improve the stability of chest bitmaps, a multi-parameter fusion method is proposed to perform pixel-level fusion of jitter corrected sequence images. Then, rough extraction of bullet hole area, energy screening and overlapping bullet holes discrimination are carried out to obtain the location distribution of bullet holes. The experimental results show that the algorithm can effectively suppress the noise caused by non-uniform illumination and random jitter, and has great ability of bullet hole extraction.

An Improved Fuzzy C-Means clustering algorithm (IU-IFCM) is proposed in this study in accordance with the characteristics of Interval Uncertain data. First, the interval data is transformed into real data composed of 2p dimension feature, which is mapped from that of p dimension feature. Second, a method for calculating sample distance, which realizes the interval sample clustering by fuzzy c-mean algorithm, is designed while considering the relationship between interval median value and interval size. Theoretical analysis and comparison experiments show that the presented algorithm surpaes the compared algorithms by more than 10% on average in terms of the Partition Coefficient (PC) and Correct Rank(CR) value. These results indicate that the algorithm presents in this study has better clustering accuracy and provides a new solution for the classification of uncertain data in current big data environments.

In recent year, the Many-objective Optimization Problems (MaOPs) have become an increasingly hot research area in evolutionary computation. However, it is still a difficult problem to achieve a good balance between convergence and diversity on solving various kinds of MaOPs. To alleviate this issue mentioned above, a Decomposition and dominance relation based many-objective Evolutionary Algorithm(DdrEA) is proposed in this paper. Firstly, the population is decomposed into numbers of sub-populations by using a set of uniform weight vectors, in which they are optimized in a cooperative manner. Then, the fitness value of solution in each sub-population is calculated by angle dominance relation and angle. Finally, elite selection strategy is performed according to its corresponding fitness value. That is, in each subspace, the solution with the smallest fitness value is selected as the elite solution to enter the next generation. Comparing with several high-dimensional and multi-objective evolutionary algorithms (NSGA-II/AD, RVEA, MOMBI-II), the experimental results show that the performance of the proposed algorithm DdrEA is better than that of the comparison algorithm, and the convergence and diversity of the population can be effectively balanced.

Many optimization problems develop into high-dimensional large-scale optimization problems in the process of the development of artificial intelligence. Although the high-dimensional problem can avoid the algorithm falling into local optimum, it has no advantage in convergence speed and time feasibility. Therefore, the natural computing method for Nonlinear Dimension Reduction (NDR) is proposed. This strategy does not depend on specific algorithm and has universality. In this method, the initialized N individuals are regarded as a matrix of N rows and D columns, and then the maximum linear independent group is calculated for the column vector of the matrix, so as to reduce the redundancy of the matrix and reduce the dimension. In this process, since any remaining column vector group can be represented by the maximum linearly independent group, a random coefficient is applied to the maximum linearly independent group to maintain the diversity and integrity of the population. The standard genetic algorithm and particle swarm optimization using NDR strategy compare with Particle Swarm Optimization (PSO), Genetic Algorithm (GA) and the four mainstream algorithms for dimension optimization. Experiments show that the improved algorithm has strong global convergence ability and better time complexity for most standard test functions.

Chronic cardiovascular diseases such as arrhythmia seriously affect human health. The automatic classification of ElectroCardioGram(ECG) signals can effectively improve the diagnostic efficiency of such diseases and reduce labor costs. To tackle this problem, an improved Long-Short Term Memory (LSTM) method is proposed to achieve automatic classification of one dimensional ECG signals. Firstly, deep Convolutional Neural Network (CNN) is designed to deeply encode the ECG signal, and ECG signal morphological features are extracted. Secondly, the LSTM classification network is used to realize automatic classification of arrhythmia of ECG signal features. Experimental studies based on the MIT-BIH arrhythmia database show that the training duration is significantly shortened and more than 99.2% classification accuracy is obtained. Sensitivity and other evaluation parameters are improved to meet the real-time and efficient requirements for automatic classification of ECG signals.

In view of the great difference between classical trajectory and real-time trajectory, a robust trajectory similarity measurement method is proposed based on the longest common subsequence theory. Firstly, the distance between point and line segment is used to judge whether the point of classical trajectory is consistent with the line segment of real-time trajectory. Secondly, the longest common subsequence length between classical trajectory and real-time trajectory is calculated by using the improved multi-to-one longest common subsequence algorithm. Finally, the ratio of the longest common subsequence length to the number of points of the classical trajectory is taken as the similarity between the classical trajectory and the real-time trajectory. Experiments show that the algorithm is robust and can effectively solve the problem of similarity measurement between the classical trajectories and real-time trajectories.

The multipath delay for different propagation mode is 0.5～2.0 ms, and the multipath delay for the same propagation mode is analyzed. Taking into account the earth magnetic field effects, the refractive index of High frequency propagation in ionosphere is combined with ray tracing, and then a new numerical iteration algorithm is given. The multipath delay caused by ionosphere dispersion is analyzed by numerical method, and the simulation is realized. Thus the analogue bandwidth of wideband communication for high frequency should be 48 kHz.

In the passive tracking using acoustic arrays, continuous and stable tracking of targets is important. In complex underwater environments, there are inevitably many trajectory interruptions, outliers, interference and target azimuth crossings in the bearing detection results, due to interference, noise, and arrays aperture limitations. In this paper, a multi-target passive tracking algorithm based on unmanned underwater vehicle is proposed. The particle sampling prediction method based on the motion information of the vehicle is used to perform the interruption prediction. The observation threshold setting method based on the motion information of the vehicle is used to adaptively set the tracking threshold. The block association tracking method is used for association of trajectory break and azimuth cross. The experimental results show that the proposed algorithm achieves correct multi-target tracking.

The space-borne Synthetic Aperture Radar (SAR) sparse flight three-dimensional (3-D) imaging technology through the multiple observations in cross-track direction obtains the 3-D spatial distribution of the observed scene. In this paper, the orbit distribution of single satellite SAR sparse flight is given. In order to shorten effectively the satellite revisit time, the formation of double star SAR orbit distribution is given. The corresponding cross-track equivalent aperture length is 20 km. A sparse 3-D imaging method based on interferometry and compressed sensing is proposed. The referential complex image is formed by using part of the echoes of the sparse flight, and the SAR 3-D image signals which are to be reconstruct are processed by interferometry. This method makes the signal sparse in the frequency domain. Under the large orbit distribution range, the frequency domain range direction and cross-track linear measurement matrix is established, which is beneficial to the Compressed Sensing(CS) theory to solve jointly the image frequency spectrum under sparse representation, and avoid the echoes coupling between the range and cross-track direction. Inversely transforming the resulting spectrum into the spatial domain, the reconstruction result can be obtained. Simulation results show that under the condition of sparse sampling rate of 74.4%, the imaging  performance of the proposed method is still comparable to that of full sampling.

The micro-Doppler modulation generated by the rotor rotation of UAV can reflect the micro-movement characteristics of such targets. Accurately estimating the rotor length and rotation frequency of the UAV is of great significance for UAV detection and recognition. In this paper, a method for estimating micro-movement parameters of multi-rotor UAV based on Concentration of Time-Frequency (CTF) is proposed for FMCW radar system. The mapping relationship between dynamic parameters of UAV rotor and signal parameters of micro-Doppler component is deduced. Based on time-frequency concentration index in time-frequency rotation domain, the discrimination of micro-motion components is improved. Compared with the traditional methods, the proposed method can improve the accuracy of multi-component micro-Doppler parameter. Furthermore, it has good robustness in low SNR. The validity of the method is verified by simulation and field test.

As the azimuth angular resolution is limited by the antenna length in automotive radars, a novel imaging approach for improving azimuth angular resolution of automotive radars is proposed based on multi-beam real-aperture radar images combination processing. Firstly, the antenna beam of the phased array antenna is electronically scanned to obtain forward-looking real-aperture radar images. Afterwards, multiple real-aperture radar images are coherent accumulated according to the imaging geometry of automotive radar to improve azimuth angular resolution. Simulation results validate the proposed imaging approach to improve the azimuth angular resolution of automotive radar.

In Specific Emitter Identification (SEI), the stability of individual features and final correct identification rate are always declined due to the influence of the primary signal with high energy on the individual features. To solve the problem above, a primary signal suppression algorithm based on synchrosqueezed wavelet transform is exploited for specific emitter identification in this paper. Firstly, a denoising method based on stationary wavelet transform is applied to preprocess the noised signal; Then, the detection and suppression of the primary signal from time-frequency distribution are developed, where root mean square error and Pearson correlation coefficient are used as numerical indicators to measure the effectiveness of the proposed primary signal suppression algorithm; Finally, a feature extraction based on box-counting dimension and a classification based on support vector machine are exploited to verify the identification performance. The simulation results show that the correct identification rate of SEI using the proposed primary signal suppression outperforms the conventional SEI with 10%, which proves the practical improvement of the proposed primary signal suppression algorithm on specific emitter identification.

When using the nonlinear Contrast Source Inversion (CSI) algorithm to solve the electromagnetic inverse scattering problem, each iteration involves finding the differential of the dissolution radiation field data about the contrast source and the total field, i.e., the Jacobi matrix. the solution of the matrix leads to the problem of large computational cost and slow convergence speed of the algorithm. in this paper, a Contrast Source Inversion algorithm based on Stochastic Average Gradient descent (SAG-CSI) is used instead of the original full gradient alternating Conjugate Gradient algorithm to reconstruct the spatial distribution information of the dielectric constant of the dielectric target under the CSI framework. the method only needs to calculate the gradient information of the randomly selected part of the measurement data in the objective function in each iteration, while the objective function keeps the gradient information of the unscented measurement data, and the optimal value of the objective function is solved together with the above two parts of the gradient information. The simulation results show that the proposed method reduces the computational cost and improves the convergence speed of the algorithm when compared with the traditional CSI method.

In order to reveal the complex electromagnetic environment effects mechanism of communication radio, the blocking effect of single-frequency and out-of-band dual-frequency electromagnetic radiation for the ultra-short wave digital communication radio is experimentally studied by irradiation method. The rule of single-frequency electromagnetic radiation effect and the susceptible bandwidth are determined. The experimental data show that the tested radio is 9～23 dB more susceptible to out-of-band dual-frequency third-order intermodulation blocking than single-frequency electromagnetic radiation blocking. A sensitive phenomenon of dual-frequency electromagnetic radiation which can neither be explained by the dual-frequency non-intermodulation superposition mechanism nor by the third-order intermodulation mechanism has been found in the experiment.