The complexity of the pre-processing phase of the cubic attack grows exponentially with the number of output bit algebras, and the difficulty of finding an effective cube set increases. In this paper, the algorithm of preprocessing stage in cubic attack is improved. In the cube set search, from random search to target search, a new target search optimization algorithm is designed to optimize the computational complexity of the preprocessing stage. In turn, the offline phase time complexity is significantly reduced. The improved cubic attack combined with the side-channel method is applied to the MIBS block cipher algorithm. The algorithm characteristics of MIBS are analyzed from the perspective of side-channel attack. The leak location is selected in the third round, and the overdetermined linear equations from initial key and output bit are established, which can directly recover 33bit key. Then the 6bit key can be recovered by quadric-detecting. The amount of plaintext required is 2^21.64, time complexity is 2²⁵. This result is greatly improved compared with the existing results, the number of keys recovered is increased, and the time complexity of the online phase is reduced.

The searchable encryption technology enables users to encrypt data and store it in the cloud, and can directly retrieve ciphertext data. Most existing searchable encryption schemes are single-to-single mode, and the searchable encryption scheme in some multi-user environments is based on public key cryptography or identity-based public key cryptosystem. Such schemes have certificate management and key escrow issues and scheme are vulnerable to suffer internal keyword guessing attacks. Public key authentication encryption and proxy re-encryption technology are combined, and an efficient certificateless authentication searchable encryption scheme is proposed for multi-user environment. The scheme uses proxy re-encryption technology to re-encrypt portion of ciphertexts, so that authorized users can generate trapdoor with the keywords to query ciphertext. In the random oracle model, the scheme is proved that it has the ability to resist the internal keyword guessing of two type attackers in the certificateless public key cryptosystem, and the calculation and communication efficiency of the scheme is better than the similar scheme.

In order to diagnose fault units in the large-scale multiprocessor systems more quickly and accurately, a system-level fault diagnosis algorithm—FireWorks Algorithm-Back Propagation Fault Diagnosis (FWA-BPFD) based on fireworks algorithm and Back Propagation(BP) neural network is proposed. Firstly, two population strategy, cooperative operator and optimal operator are introduced into fireworks algorithm. A new fitness function is designed, and the mutation operator, mapping rule and selection strategy are optimized. Then, the optimization process of weight and threshold value in BP neural network is optimized by the self-regulating mechanism of global and local searching ability of fireworks algorithm. Simulation results show that compared with other algorithms, this algorithm not only reduces the number of iterations and training time, but also improves the accuracy of diagnosis.

For the fixed and limited number of OpenFlow protocol matching fields, and the lack of effective forwarding verification mechanism for data packet forwarding in the Software-Defined Networking (SDN), a SDN packet forwarding verification mechanism based on programmable data plane is proposed. By adding a cipher identification to the data packet, the P4 forwarding device joins the OpenFlow-based SDN network to control accurately and sample network traffic flow without affecting the normal forwarding of the data flow. The controller verifies the integrity of the sampled packet, and sends flow rules to the OpenFlow forwarding device to control the abnormal data flow such as malicious tampering and forgery. Finally, the forwarding verification prototype and simulation network based on P4 forwarding device and Open vSwitch forwarding device are constructed and tested. The experimental results show that the mechanism can effectively detect the forwarding abnormal behaviors such as packet tampering and forgery. Compared with similar verification mechanisms, in the case of the same security verification processing overhead, it can achieve more fine-grained flow precise control sampling and lower forwarding delay.

Due to the limitation of energy and bandwidth in Wireless Sensor Networks(WSN), the direct transmission of analog signals in the network is greatly restricted. Therefore, quantization of analog signals is an important means to save network energy and ensure effective bandwidth. To this end, based on the principle of minimum absolute mean reconstruction error a network quantization and energy optimization method is designed in this paper. Firstly, for single sensor, the optimal quantization bit number is derived under the condition of fixed energy and the optimal energy distribution is derived under the condition of fixed quantization bit number. Secondly, on the basis of single sensor, the optimal quantization bit number and optimal energy allocation are further deduced in multi-sensor case. In both cases, the sensor measurement noise and channel fading loss are considered. Finally, the numerical simulation results show that the proposed method is correct and better than the equal energy distribution.

Vehicular Ad hoc NETworks (VANETs) which is an important part of smart cities are large networks that organize wireless communication and information exchange between vehicles and X (X: cars, roads, pedestrians, and the Internet). The security and efficiency of the message authentication algorithm are crucial to the VANETs. After analyzing the security shortage of Wang Daxing et al VANETs message authentication scheme, an improved provable secure certificateless aggregation signature scheme for VANETs is proposed. The scheme constructs a secure certificateless aggregation authentication scheme by using Elliptic Curve Cryptography (ECC) and reduces the complexity of the cryptographic operation process, while achieving user’s conditional privacy protection. Rigid security analysis proves that the scheme satisfies the security requirements of VANETs. The performance analysis shows the proposed scheme considerably reduces the computational cost of message signature, single verification and aggregation verification algorithm, and reduces the communication cost when compared with Wang schemes.

To solve the problem of asynchronous sampling and communication delay of sensor network in space target tracking, an Asynchronous Distributed algorithm based on Information Filtering (ADIF) is proposed. First, local state information and measurement information with sampling time is transmitted between local sensor and adjacent nodes in a certain topology structure. Then, the local sensor sorts the received asynchronous information by time, and ADIF algorithm is used to calculate the target state respectively. This method is simple to implement, the frequency of communication between sensors is small, and it supports the real-time change of network topology, which is suitable for multi-target tracking. In this paper, single target and multi-target tracking are simulated respectively. The results show that the algorithm can effectively solve the problem of asynchronous sensor filtering, and the distributed filtering accuracy converges to the centralized result.

There are some limitations in the existing metric methods for measuring node influence. A measurement method of node influence with three-level neighbors is proposed, which is based on the principle of three-degree influence, and considering the appropriate level of local measurement and the scalability of the large-scale network. Firstly, the neighbors with propagation attenuation characteristics in the second and third level of a node are regarded as a whole, which is used to measure the influence of the node. Then, an algorithm for measure called Three-level Influence Measurement (TIM) is proposed. Finally, in order to validate the effectiveness of the algorithm, the experiments on three datasets are conducted by using susceptible-infected-recovered model and independent cascade model. The experimental results show that the proposed algorithm is superior in consistency of influence, discrimination, sorting performance and other evaluation indexes. Furthermore, the TIM is applied to effectively solve the problem of maximizing influence.

Wireless Local Area Network (WLAN) indoor intrusion detection technique is one of the current research hotspots in the field of intelligent detection, but the conventional database construction based intrusion detection technique does not consider the time-variant property of WLAN signal in the complicated indoor environment, which results in the low robustness of WLAN indoor intrusion detection system. To address this problem, a Multiple Kernel Maximum Mean Discrepancy (MKMMD) transfer learning based WLAN indoor intrusion detection approach is proposed. First of all, the offline labeled and online pseudo-labeled Received Signal Strength (RSS) features are used to construct source and target domains respectively. Second, the optimal transfer matrix is constructed to minimize the MKMMD of the joint distributions of RSS features in source and target domains. Third, a classifier trained from the transferred RSS features and the corresponding labels in source domain is used to classify the transferred RSS features in target domain, and meanwhile the label set corresponding to target domain is obtained. Finally, the label set corresponding to target domain is updated in an iterative manner until the proposed algorithm converges, and then the intrusion detection in target environment is achieved. The experimental results indicate that the proposed approach is able to preserve high detection accuracy as well as overcome the impact of time-variant signal property on the detection performance.

Due to the influence of many factors such as multipath and multi-source, the traditional indoor localization algorithms based on Bluetooth signal strength have low performance in accuracy and stability. In order to solve the location problem in complex indoor environment based on Bluetooth signal, an indoor localization algorithm based on low-cost array antenna is developed. The algorithm utilizes single-channel using switch-antenna polarization sensitive array to sample Bluetooth signal, then combines the accurate array manifold measured in dark room and the algorithm of Polarized Fast Converging Sparse Bayesian Learning (P-FCSBL) to estimate the source’s angle, and finally gets the target location by angle. This algorithm makes full use of polarization information and angle information to separate target and multipath signal, and simultaneous sampling of one source ensures estimation stability. Finally, the effectiveness of the method is verified by the real data.

Indoor positioning technology based on Channel State Information (CSI) receives much attention in recent years. The existing indoor positioning solution is continuously innovative and improved in terms of deployment implementation and positioning accuracy. This paper proposes a passive one-transmitter two-receivers fingerprint indoor positioning system. The CSI data is collected by two fixed receiving end-devices. In the signal preprocessing stage, the CSI amplitude is singular value removed and low pass filtered, and the CSI phase is corrected by a linear fitting method, and the CSI amplitude and phase information obtained by the two receiving ends are collectively used as fingerprint samples. The fingerprint samples are finally trained through the fully connected neural network, and matched with the collected real-time data. Experiments show that the matching recognition rate reaches 98% by using two receivers and the combination of amplitude and phase positioning, and the positioning accuracy is 0.69 m. It proves that the system can accurately and effectively achieve indoor positioning.

Polar Sine Transform (PST) is used to detect Copy-move forgeries in the paper, and the image to be detected is transformed into gray scale image and feature extraction is carried out by PST. Improved PatchMatch, a fast approximate nearest neighbor search algorithm, is used to match feature descriptors to overcome the problem of long time consuming caused by matching global descriptors. Experiments show that the proposed method is not only effective for linear Copy-move forgeries and rotation interference forgeries, but also robust to noise and JPEG compression interference forgeries. Finally, the experimental results of synthetic interference forgeries show that the accuracy can reach 98.0% when the synthetic forgeries range is small.

The label-specific features learning avoids the same features prediction for all class labels, it is a kind of framework for extracting the specific features of each label for classification, so it is widely used in multi-label learning. For the problems of large label dimension and unbalanced label distribution density, the existing multi-label learning algorithm based on label-specific features has larger time consumption and lower classification accuracy. In order to improve the performance of classification, a Group-Label-Specific Features Learning method based on Label-Density Classification Margin (GLSFL-LDCM) is proposed. Firstly, the cosine similarity is used to construct the label correlation matrix, and the class labels are grouped by spectral clustering to extract the label-specific features of each label group to reduce the time consumption for calculating the label-specific features of all class labels. Then, the density of each label is calculated to update the label space matrix, the label-density information is added to the original label space. The classification margin between the positive and negative labels is expanded, thus the imbalance label distribution density problem is effectively solved by the method of label-density classification margin. Finally, the final classification model is obtained by inputting the group-label-specific features and the label-density matrix into the extreme learning machine. The comparison experiment results verify fully the feasibility and stability of the proposed algorithm.

Deep learning has shown excellent performance in the field of artificial intelligence. In the supervised identification task, deep learning algorithms can achieve unprecedented recognition accuracy by training massive tagged data. However, owing to the high cost of labeling massive data and the difficulty of obtaining massive data of rare categories, it is still a serious problem how to identify unknown class that is rarely or never seen during training. In view of this problem, the researches of Zero-Shot Learning (ZSL) in recent years is reviewed and illustrated from the aspects of research background, model analysis, data set introduction and performance analysis in this article. Some solutions of mainstream problem and prospects of future research are provided. Meanwhile, the current technical problems of ZSL is analyzed, which can offer some references to beginners and researchers of ZSL.

It is difficult for current salient object detection algorithms to reach a good balance performance between accuracy and efficiency. To solve this problem, a deep convolutional network for saliency object detection with balanced accuracy and high efficiency is produced. First, through replacing the traditional convolution with the decomposed convolution, the computational complexity is greatly reduced and the detection efficiency of the model is improved. Second, in order to make better use of the characteristics of different scales, sparse cross-layer connection structure and multi-scale fusion structure are adopted to improve the detection precision. A wide range of evaluations show that compared with the existing methods, the proposed algorithm achieves the leading performance in efficiency and accuracy.

A high-resolution controllable magnification method for visual saliency object based on virtual optics is proposed in this paper. The original image is placed on the virtual object plane. Firstly, the diffractive wave of the original image on the virtual diffraction plane is obtained by inverse diffraction calculation, and then the forward diffraction calculation is carried out after the virtual diffraction wave is irradiated by spherical wave. The original images with different magnification can be reconstructed by changing the position of the observation plane. The simulation results show that compared with the general interpolation method, the magnified image shows a good visual perception effect, especially in the saliency region. When the degraded face image is used as the signal to be reconstructed, the significant areas such as eyes and nose are clearer than the general method. The local salient region in the original image is segmented by the level set method combined with salient map, and the magnification and contour extraction are performed. The contours show good smoothness.

Image-to-image translation is a method to convert images in different domains. With the rapid development of the Generative Adversarial Network(GAN) in deep learning, GAN applications are increasingly concerned in the field of image-to-image translation. However, classical algorithms have disadvantages that the paired training data is difficult to obtain and the convert effect of generation image is poor. An improved Cycle-consistent Generative Adversarial Network(CycleGAN++) is proposed. New algorithm removes the loop network, and cascades the prior information of the target domain and the source domain in the image generation stage, The loss function is optimized as well, using classification loss instead of cycle consistency loss, realizing image-to-image translation without training data mapping. The evaluation of experiments on the CelebA and Cityscapes dataset show that new method can reach higher precision under the two classical criteria—Amazon Mechanical Turk perceptual studies(AMT perceptual studies) and Full-Convolutional Network score(FCN score), than the classical algorithms such as CycleGAN, IcGAN, CoGAN, and DIAT.

Feature subspace learning is a critical technique in image recognition and classification tasks. Conventional feature subspace learning methods include two main problems. One is how to preserve the local structures and discrimination when the samples are projected into the learned subspace. The other hand when the data are corrupted with noise, the conventional learning models usually do not work well. To solve the two problems, a discriminative feature learning method is proposed based on Low Rank Representation (LRR). The novel method includes three main contributions. It explores the local structures among samples via low rank representation, and the representation coefficients are used as the similarity measurement to preserve the local neighborhood existed in the samples; To improve the anti-noise performance, a discriminative learning item is constructed from the recovered samples via low rank representation, which can enhance the discrimination and robustness simultaneously; An iterative numerical scheme is developed with alternating optimization, and the convergence can be guaranteed effectively. Extensive experimental results on several visual datasets demonstrate that the proposed method outperforms conventional feature learning methods on both of accuracy and robustness.

In order to improve the correlation between signal samplings and reduce the influence of noise on sensing performance, a spectrum sensing algorithm based on signal envelope autocorrelation matrix is proposed in the paper. Firstly, the sampling signals are intercepted at equal intervals, the signal autocorrelations are calculated by means of the adjacent interval samples, and an approximate autocorrelation matrix is constructed. Secondly, the statistic is constructed according to the properties of the sub-diagonal elements of the matrix. The detection probability distribution function and the false alarm probability distribution function of the statistic are calculated respectively. The detection performances of the spectrum sensing algorithm are analyzed. The algorithm optimizes the calculation of signal correlation and reduces the impact of noise on detection performance. Finally, the effects of different parameters on detection probability and false alarm probability are verified by simulation, and some measures are proposed to improve detection performance.

A subchannel matching method based on bilateral matching and a power allocation algorithm based on Stackelberg game are proposed for two-tier Non-Orthogonal Multiple Access (NOMA) network. Firstly, the resource optimization problem is decomposed into two subproblems—sub-channel matching and power allocation. In the power allocation, the macro base station layer and small base station layer are regarded as the leader and followers in the Stackelberg game. Then, the non-convex optimization problem is converted into a way to be easily solved, and the power allocation of the both layers are obtained respectively. Finally, the global power allocation scheme of the system is obtained by using Stackelberg game. The simulation results show that the proposed resource optimization algorithms can effectively improve the energy efficiency of the two-tier NOMA system.

For online dynamic radio resources optimization for network slices in Heterogeneous Cloud Raido Access Network (H-CRAN), by comprehensively considering traffic admission control, congestion control, resource allocation and reuse, the problem is formulated as a stochastic optimization programming which maximizes network average total throughput subject to Base Station (BS) transmit power, system stability, Quality of Service (QoS) requirements of different slices and resource allocation constraints. Then, a joint congestion control and resource allocation dynamic scheduling algorithm is proposed which will dynamically allocate resources to users in network slices with distinct performance requirements within each resource scheduling time slot. The simulation results show that the proposed algorithm can improve the network overall throughput while satisfying the QoS requirement of each slice user and maintaining network stability. Besides, it could also flexibly strike a dynamic balance between delay and throughput by simply tuning an introduced control parameter.

Utilizing the characteristics of the wireless channel to achieve secure transmission of information through physical layer technology is a way to realize security communications. The time-reversed transmission has natural anti-jamming and anti-eavesdropping capability due to its unique spatial and temporal focusing property, so a good secrecy performance can be obtained even when the transmitter is equipped with single transmitting antenna. This paper studies the optimization of the transmit filter impulse response in a two-user time-reversed downlink multiple access secure transmission system. The joint optimization problem of two transmitting filters is transformed into the independent optimization problem of each filter based on reciprocity principles. This problem is further converted into the problem of finding the largest eigenvalue and its corresponding eigenvector which is solved by iterative algorithm. The simulation results show that by optimizing the pre-filter for sum secrecy rate, the system's sum secrecy rate is promoted and is obviously higher than that of the conventional time-reversed pre-processing filter system and direct transmission system.

Cyclic Redundancy Check (CRC) is used in cascade with channel coding to improve the convergence of the decoding. In the new generation of wireless communication systems, such as 5G, both code length and code rate are diverse. To improve the decoding efficiency of cascaded systems, a CRC parallel algorithm with variable computing width is proposed in this paper. Based on the existing fixed bit-width parallel algorithm, this algorithm combines the parallel calculation of feedback data and input data in the formula recursive method, realizing a highly parallel CRC check architecture with variable bit-width CRC calculation. Compared with the existing parallel algorithms, the merged algorithm saves the overhead of circuit resources. When the bit-width is fixed, the resource saving effect is obvious, and at the same time, the feedback delay is also optimized by nearly 50%. When the bit-width is variable, the use of resources is also optimized accordingly.

Spoofing misleads the receiver to generate the wrong position information by trans-mitting signals similar to authentic satellite signals, which has great harm. In this paper, a single-antenna spoofing mitigation algorithm based on signal reconstruction is proposed for meaconing. Firstly, the carrier frequency and code phase of spoofing signal are obtained by parameter estimation method, and then the orthogonal projection matrix of spoofing signal subspace is constructed to suppress spoofing. The simulation results show that the algorithm has a good suppression effect on spoofing and ensure the receiver can locate effectively in the interference environment, and the algorithm also has lower computational complexity.

In order to reduce the sampling rate of the Traveling Wave Tube (TWT) of the Analog to Digital Converter (ADC) in the feedback loop of Digital PreDistortion (DPD), the nonlinear parameters of the power amplifier model are proved to be estimated with the undersampled output signal based on the cyclostationary of digital modulation signal. The output signal similar to high sampling rate can be obtained by combining the nonlinear parameters of the power amplifier model with the input signal. The DPD of the power amplifier is implemented through indirect learning architecture. To validate the method, a 55 W X-band Traveling Wave Tube Amplifier (TWTA) is driven by a 20 MHz LTE signal. The sampling rate of ADC in the DPD feedback loop is reduced from 61.44 Msps to 6.144 Msps and 3.072 Msps, but the linearization effect has little change, which shows the validation of the undersampling method.

For solving the problem of the synthesis of sparse rectangular planar arrays with multiple constraints, this paper proposes a Dynamic Parameters Differential Evolution (DPDE) based algorithm. Firstly, to improve searching efficiency and accuracy of Differential Evolution (DE), the proposed method introduces dynamically changing strategies to the scaling factor and the crossover probability of the traditional Differential Evolution algorithm. Secondly, a modified matrix mapping method and the redefinition of mapping principles are presented to make up the defects of strong randomness and low accuracy in existing methods. Finally, simulation experiments of antenna arrays are performed to validate the effectiveness of the proposed method, and the results demonstrate that the proposed method performs out the existing methods in the respect of reducing peak sidelobe level of antenna arrays.

For the problem of co-frequency base station interference in passive radar based on Long Term Evolution (LTE) signal, an algorithm based on blind source separation using second order statistics is proposed. The presented algorithm is based on convolution mixed model, and achieves the minimum correlation among separated signals through multi-channel Least-Mean-Square (LMS) algorithm. Without statistical correlation among the signals of each transmitting base station, the separation of the observed signals is completed when the separated signals achieve the minimum correlation. On this basis, the traditional signal processing for passive radar is improved. The steps of separating co-frequency interference clutter consisting of both direct-path and multipath clutter are added, which can suppress the clutter interference of co-channel base station. Simulation and analysis verify the effectiveness of the algorithm. The algorithm provides a reference for data processing of passive radar based on LTE signal.

A robust spatial sign transform-based maximum likelihood method for low-elevation target altitude measurement is proposed in the presence of the non-Gaussian diffuse multipath component for Very High Frequency (VHF) radar. The spatial sign transform is implemented to the antenna array snapshots, reducing the influence of the outliers on array covariance matrix and the low elevation estimation algorithms, followed by computing the spatial Sign Covariance Matrix(SCM). Then the application of SCM to the Maximum Likelihood method(SCM-ML) is presented on the basis of the affine equivalence and preservation of the eigenstructure for robust low elevation estimation and height finding of VHF radar. The proposed method effectively solves the non-Gaussian property of the diffuse multipath component and improves the robustness of low elevation estimation. Simulation result and real data demonstrate the robustness and validation of the SCM-ML method.