Current Issue

25 November 2023, Volume 44 Issue 11

Previous Issue   

Topics: Key Technologies for Ubiquitous Sensing and Intelligent Recognition in the Ubiquitous Internet of Things

Cloud-edge-device fusion architecture oriented to spectrum cognition and decision in low altitude intelligence network

Chao DONG, Yuqian JING, Yuben QU, Bo ZHOU, Yang HUANG, Ziye JIA, Haipeng DAI, Qihui WU

2023, 44(11):  1-12.  doi:10.11959/j.issn.1000-436x.2023228

Asbtract ( 264 )   HTML ( 74)   PDF (6188KB) ( 337 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

A new architecture for the typical application scenario of low altitude intelligence network spectrum cognition and decision was introduced.Firstly, the challenges of spectrum security management and efficient sharing in low altitude intelligence network were analyzed, and the key scientific problems that need to be solved were summarized.Secondly, to support the spectrum management and sharing of low altitude intelligence network through deep integration of cloud, edge, and device, the cloud-edge-device fusion architecture for low altitude intelligence network spectrum cognition and decision was proposed.Then, the key technologies such as fast and accurate spectrum recognition and agile adaptive decision making based on cloud-edge-device fusion were discussed.Finally, the future research direction of the architecture was introduced.

Multi-sensing node convolution fusion identity recognition algorithm for radio digital twin

Guofeng WEI, Guoru DING, Yutao JIAO, Yitao XU, Daoxing GUO, Peng TANG

2023, 44(11):  13-24.  doi:10.11959/j.issn.1000-436x.2023227

Asbtract ( 162 )   HTML ( 39)   PDF (5863KB) ( 223 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

Electromagnetic space is an important link to empower and coordinate sea, land, air, space and network.Electromagnetic target recognition provides important radio target identity information for the twin construction of electromagnetic space, so that it can describe and depict the identity situation of electromagnetic targets in digital space.However, a single sensing node is vulnerable to interference, and its recognition performance is limited.Wrong recognition results will provide radio digital twin with conflicting identity information.Therefore, based on the requirements of radio digital twin in electromagnetic space, a radio target recognition framework for radio digital twin was constructed and a multi-sensing node convolution neural network individual identity fusion recognition algorithm was proposed.Compared with the nearest single sensing node, the recognition performance is improved by 6.29% by deploying the multi-node recognition network in the actual scene, which provides more accurate individual identity information.

Research progress on electromagnetic spectrum multidimensional situation compressed mapping technology

Feng SHEN, Guoru DING, Jie LI, Bo ZHOU, Qihui WU

2023, 44(11):  25-42.  doi:10.11959/j.issn.1000-436x.2023174

Asbtract ( 133 )   HTML ( 27)   PDF (2440KB) ( 269 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

In the increasingly complex electromagnetic spectrum environment, accurately obtaining comprehensive spectrum situation is a crucial prerequisite for making precise spectrum decisions.First, the spectrum mapping was introduced and compared with spectrum sensing.Then, an in-depth review of existing spectrum situation generation methods was conducted.Next, multidimensional spectrum situation compressed mapping in the face of challenges such as heterogeneity, large scale missing data, time variability and environmental complexity was proposed.It effectively compensated for the incompleteness of the spectrum mapping framework caused by ignoring the spectrum situation sensing process in traditional spectrum situation generation methods.This could further provide more accurate guidance for enhancing spectrum utilization efficiency, strengthening spectrum security maintenance, and intensifying electromagnetic warfare decision-making.Lastly, the future development trends of spectrum compressed mapping were discussed.

Prototype verification for integrated sensing and communications:current status and development trends

Jie YANG, Yixuan HUANG, Tao DU, Hang QUE, Shuqiang XIA, Shi JIN

2023, 44(11):  43-54.  doi:10.11959/j.issn.1000-436x.2023205

Asbtract ( 240 )   HTML ( 51)   PDF (1731KB) ( 439 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

Considering the practical challenges faced by the commercialization of integrated sensing and communication (ISAC), the implementation progress and future trends of ISAC were elaborated and summarized.First, the research progress and empirical performance of single node-based ISAC prototype systems were summarized, including positioning, environmental mapping, imaging, waveform design, and beam tracking.Next, the actual gain of multi-nodes cooperation-based ISAC prototype systems in the scenarios of the Internet of vehicles and the Internet of things were analyzed.Then, focusing on cross-sensor-based ISAC prototype systems, the advantages and disadvantages of each kind of sensor and the experimental fusion performance were overviewed.Finally, the future trends in the system implementation and technical validation of ISAC were highlighted, addressing the challenges encountered in existing ISAC prototype systems.

Topics: Distributed Edge Intelligence for Complex Environments

Live video transmission optimization mechanism based on edge intelligence in high client-density environment

Xiaodan GU, Wenjia WU, Zhen LING

2023, 44(11):  55-66.  doi:10.11959/j.issn.1000-436x.2023232

Asbtract ( 93 )   HTML ( 14)   PDF (2768KB) ( 211 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

The traditional live video transmission optimization mechanism is deployed on the server side, which cannot quickly respond to the dynamic changes of the user’s wireless network environment.To address this problem, a live video transmission optimization mechanism based on edge intelligence called S-Edge was proposed.It was deployed on the OpenWrt-based wireless access point, and comprehensively utilized the wireless channel state information such as airtime utilization and signal-to-noise ratio to make intelligent decisions on terminal priority and transmission rate based on fuzzy logic theory.Furthermore, the active queue management with hierarchical token bucket and service demand-driven wireless transmission rate adaptive control technologies were introduced to realize the real-time scheduling of live video data.In order to verify the effectiveness and performance of the proposed mechanism, a high client-density environment was built through user clusters based on multi-radio interfaces in the real-world scenario.Experimental results show that S-Edge can significantly reduce the average delay and packet loss rate, which meets QoS requirements of live video transmission services in the high client-density environment.

Resource sharing and incentive mechanism for multi-access edge computing networks

Liangjun SONG, Gang SUN, Jian SUN, Hongfang YU

2023, 44(11):  67-78.  doi:10.11959/j.issn.1000-436x.2023210

Asbtract ( 125 )   HTML ( 24)   PDF (1217KB) ( 147 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

A tripartite incentive-enabled edge device resource sharing algorithm was proposed to manage and allocate the shared resources for the “cloud-edge-end” three-tier edge network architecture.Considering the single-point network congestion, network-wide coverage constraint, limited resource constraint and transmission cost constraint, the edge devices were abstracted into different functional nodes to design a convex envelope-based shared node (CESN) selection algorithm to achieve a full-coverage, low-energy shared resource deployment scheme.Secondly, to address the additional cost problem caused by allocating shared resources on edge devices, a tripartite sharing incentive mechanism was designed based on the shared resource leasing model, which guaranteed the long-term and stable participation of edge devices in the sharing mechanism.The simulation results show that the proposed algorithm significantly improves the system energy consumption and node revenue compared to the existing algorithms.

Client grouping and time-sharing scheduling for asynchronous federated learning in heterogeneous edge computing environment

Qianpiao MA, Qingmin JIA, Jianchun LIU, Hongli XU, Renchao XIE, Tao HUANG

2023, 44(11):  79-93.  doi:10.11959/j.issn.1000-436x.2023196

Asbtract ( 239 )   HTML ( 55)   PDF (2317KB) ( 352 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

To overcome the three key challenges of federated learning in heterogeneous edge computing, i.e., edge heterogeneity, data Non-IID, and communication resource constraints, a grouping asynchronous federated learning (FedGA) mechanism was proposed.Edge nodes were divided into multiple groups, each of which performed global updated asynchronously with the global model, while edge nodes within a group communicate with the parameter server through time-sharing communication.Theoretical analysis established a quantitative relationship between the convergence bound of FedGA and the data distribution among the groups.A time-sharing scheduling magic mirror method (MMM) was proposed to optimize the completion time of a single round of model updating within a group.Based on both the theoretical analysis for FedGA and MMM, an effective grouping algorithm was designed for minimizing the overall training completion time.Experimental results demonstrate that the proposed FedGA and MMM can reduce model training time by 30.1%～87.4% compared to the existing state-of-the-art methods.

Survey on model inversion attack and defense in federated learning

Dong WANG, Qianqian QIN, Kaitian GUO, Rongke LIU, Weipeng YAN, Yizhi REN, Qingcai LUO, Yanzhao SHEN

2023, 44(11):  94-109.  doi:10.11959/j.issn.1000-436x.2023209

Asbtract ( 341 )   HTML ( 60)   PDF (4179KB) ( 418 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

As a distributed machine learning technology, federated learning can solve the problem of data islands.However, because machine learning models will unconsciously remember training data, model parameters and global models uploaded by participants will suffer various privacy attacks.A systematic summary of existing attack methods was conducted for model inversion attacks in privacy attacks.Firstly, the theoretical framework of model inversion attack was summarized and analyzed in detail.Then, existing attack methods from the perspective of threat models were summarized, analyzed and compared.Then, the defense strategies of different technology types were summarized and compared.Finally, the commonly used evaluation criteria and datasets were summarized for inversion attack of existing models, and the main challenges and future research directions were summarized for inversion attack of models.

Edge intelligence-assisted routing protocol for Internet of vehicles via reinforcement learning

Bingyi LIU, Yuhao LIU, Weizhen HAN, Zhenchang XIA, Libing WU, Shengwu XIONG

2023, 44(11):  110-119.  doi:10.11959/j.issn.1000-436x.2023187

Asbtract ( 139 )   HTML ( 26)   PDF (1266KB) ( 207 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

To achieve a highly reliable and adaptive packet routing protocol in a complex urban Internet of vehicles, an end-edge-cloud edge intelligence architecture was proposed which consisted of an end user layer, an edge collaboration layer, and a cloud computing layer.Based on the proposed edge intelligence architecture, an packet routing protocol based on multi-intelligent reinforcement learning technologies was designed.The experimental results show that the proposed protocol could significantly improve the transmission delay and the packet reception rate in the interval of 29.65%～44.06% and 17.08%～25.38% compared to the state-of-the-art transmission mechanism for emergency data (TMED), intersection fog-based distributed routing protocol (IDR), and double deep Q-net based routing protocol (DRP).

Pilot spoofing detection algorithm for edge nodes based on heterogeneous pilot energy estimation

Shiguo WANG, Shujuan TIAN, Qingyong DENG

2023, 44(11):  120-128.  doi:10.11959/j.issn.1000-436x.2023231

Asbtract ( 60 )   HTML ( 7)   PDF (1102KB) ( 178 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

For the federated learning scenarios with edge-end cooperation, edge servers and device terminals update their models and exchange gradient parameters frequently, and hence eavesdroppers can manipulate channel estimation through pilot spoofing to intercept the transmitted information and reduce the update efficiency of federated learning model.Therefore, a pilot attack detection algorithm with heterogeneous pilot energy estimation was proposed.Firstly, a federated learning pilot attack system model was constructed after the security of pilot attacks on data transmission had been analyzed.Then, a pilot attack detection method based on random segmentation and encryption methods was proposed to detect the pilot spoofing accurately and the contaminated channel could be recovered as well.Experimental results show that the proposed algorithm is more suitable for detecting pilot attacks with low transmit power and strong concealment compared to other existing algorithms.Furthermore, the data transmission rate of edge servers is improved significantly through the precoding based on the recovered channel state information.

Papers

Optimization of task scheduling for computing reuse in computing power network

Yunxiao MA, Zhonghui WU, Zuyun XU, Lujie ZHONG, Changqiao XU

2023, 44(11):  129-142.  doi:10.11959/j.issn.1000-436x.2023206

Asbtract ( 179 )   HTML ( 22)   PDF (2304KB) ( 268 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

To cope with the challenges posed by the future explosive growth in computing power demand, computing reuse technology was introduced into the computing power network to reduce service latency and computational resource consumption by reusing the results of computational tasks.Based on this, a service federation-based context-aware online learning algorithm was proposed.First, the reuse index was designed to reduce the extra lookup latency.Then, online learning was performed based on the service federation mechanism to make computational task scheduling decisions according to contextual information and historical experience.The experimental results show that the proposed algorithm outperforms the baseline algorithms in terms of service latency and computational resource consumption.

Efficiency analysis of distributed power combining for mobile platform

Wenbo GUO, Jiaxin DU, Jian YANG, Mu YAN, Hongzhi ZHAO, Shihai SHAO

2023, 44(11):  143-150.  doi:10.11959/j.issn.1000-436x.2023219

Asbtract ( 51 )   HTML ( 7)   PDF (2284KB) ( 74 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

Aiming at the position perturbation of mobile platform, the impact of position perturbation was analyzed on distributed power combining efficiency.The model of distributed power combining with position perturbation was established, and the average far-field beampattern, complementary cumulative distribution function (CCDF) expressions and the approximate expression of 3 dB beamwidth were derived.The results show that intensified position perturbation can lead to continuous deterioration of the average far-field beampattern, peak power, 3 dB beamwidth, and CCDF of distributed power combining.When the position perturbation reaches 9% of the signal wavelength, the signal peak power of the mobile platform after distributed power combining decreases to 90% of its theoretical upper limit.

CNN-based continuous authentication scheme for vehicular digital twin

Chengzhe LAI, Xinwei ZHANG, Guanjie LI, Dong ZHENG

2023, 44(11):  151-160.  doi:10.11959/j.issn.1000-436x.2023229

Asbtract ( 112 )   HTML ( 11)   PDF (1103KB) ( 117 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

To address vehicle identity legitimacy verification issues, a continuous authentication scheme for vehicular digital twin based on convolutional neural network (CNN) was proposed.Specifically, the digital twin was used to acquire the data collected by the vehicle sensors for training the CNN deployed on the digital twin.Then, principal component analysis was performed to select appropriate typical features for the classifier.Using the features extracted by the CNN, the one-class support vector machine (OC-SVM) classifier was trained in the registration phase and the data was classified in the authentication phase, which consequently verified the current vehicle as a legitimate or malicious vehicle.Simulation results show that the proposed scheme has outstanding advantages and outperforms the existing schemes in terms of performance and accuracy.

Evaluation method of penetrating jamming effectiveness for cognitive communication protection

Zhiguo SUN, Shuo XIAO, Yijie WU, Shiming LI, Zhenduo WANG

2023, 44(11):  161-172.  doi:10.11959/j.issn.1000-436x.2023195

Asbtract ( 68 )   HTML ( 12)   PDF (2066KB) ( 122 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

In order to support the implementation of “multidimensional orthogonal configured communication protection between communication modules as well as parameters and interference” in cognitive communication countermeasures scenarios, jamming effectiveness requires module level responsiveness.A penetrating jamming effectiveness evaluation method was proposed.Firstly, the core parameters of each signal processing module were selected as evaluation parameters.Secondly, analytic hierarchy process (AHP) and entropy weight method (EWM) were used to determine the weight of indicators.The coordination problem in AHP was solved through expert weight allocation, and the weight allocation problem in EWM was solved through non-uniform mapping.Finally, the subjective and objective weights were weighed through game theory, and the jamming effectiveness was judged by the technique for order preference by similarity to ideal solution (TOPSIS).By conducting evaluation experiments on the jamming performance of the Link16 data link, taking into account multiple aspects such as jamming types, number of jamming frequencies, jamming duration, and system types.Experiment results demonstrate that the penetration-based jamming performance evaluation method not only provides a more intuitive representation of the relationship between evaluation results and metric parameters, but also possesses excellent jamming discrimination capability.

Establishment and performance analysis of RIS-assisted coded relay cooperation system based on LDPC product codes

Shunwai ZHANG, Jin WANG

2023, 44(11):  173-182.  doi:10.11959/j.issn.1000-436x.2023212

Asbtract ( 78 )   HTML ( 18)   PDF (1121KB) ( 113 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

To pursue the ultra-reliable information transmission, the system which combined the reconfigurable intelligent surface (RIS) technology and coded relay cooperation technology was investigated.The RIS-assisted coded relay cooperation system model based on low density parity check (LDPC) product codes was established.Based on LDPC product codes, the information matrix was row-encoded by the source node and column-encoded by the relay node, respectively.An efficient joint iterative decoding algorithm was proposed.The two received signals at the destination were row-decoded and column-decoded iteratively to update the logarithmic likelihood ratios of information bits, and then the decoding decision was made.The closed-form solutions of the outage probability and the channel capacity with finite codelength were theoretically derived.Theoretical analysis and simulation results show that the proposed system obviously outperforms the traditional coded relay cooperation.The more number of RIS elements, the more significant advantage can be obtained.Simulation results also demonstrate that the channel capacity with finite codelength approaches the ideal channel capacity with the codelength of LDPC product codes increasing.

Heterogeneous federated bidirectional knowledge distillation transfer semi-supervised modulation recognition

Peihan QI, Yuanlei DING, Kai YIN, Jiabo XU, Zan LI

2023, 44(11):  183-200.  doi:10.11959/j.issn.1000-436x.2023191

Asbtract ( 109 )   HTML ( 13)   PDF (5671KB) ( 144 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

The large-scale deployment and rapid development of the new generation mobile communication system underpin the widespread application of a massive and diverse range of Internet of things (IoT) devices.However, the distributed application of IoT devices results to significant disparities in private data and substantial heterogeneity in local processing models, which severely limits the aggregation capability of global intelligent model.Therefore, to tackle the challenges of data heterogeneity, model heterogeneity, and insufficient labeling faced by intelligent modulation recognition in cognitive IoT, an algorithm was proposed for heterogeneous federated bidirectional semi-supervised modulation recognition, which incorporated bidirectional knowledge distillation.In the proposed algorithm, a public pseudo dataset was generated by variational autoencoder in the cloud for supporting uplink global knowledge distillation, and adaptively sharing to the local devices for downlink heterogeneous knowledge distillation, while integrating a semi-supervised algorithm within the distillation process.The simulation results indicate that the proposed algorithm outperforms current federated learning algorithms in terms of effectiveness and applicability in the field of communication signal processing.

Radio frequency fingerprint data augmentation for indoor localization based on diffusion model

Haojun AI, Weike ZENG, Jingjie TAO, Jinying XU, Hanxiao CHANG

2023, 44(11):  201-212.  doi:10.11959/j.issn.1000-436x.2023200

Asbtract ( 112 )   HTML ( 17)   PDF (1408KB) ( 245 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

The radio frequency fingerprint indoor localization method ensures the accuracy by collecting a sufficient amount of fingerprints in the offline state to build a dense fingerprint database.A data augmentation method called FPDiffusion was proposed based on diffusion model to reduce the cost of fingerprint acquisition.Firstly, a temporal graph representation of the fingerprint sequence was constructed, the forward process of the diffusion model was accomplished by adding Gaussian noise, and a U-Net was utilized for the reverse process.The loss function of the network was designed according to the characteristics of radio frequency fingerprints.Finally, the computational process for generating dense fingerprints based on sparse fingerprints was presented.Experimental results demonstrate that FPDiffusion achieves 76% and 28% localization error reduction on K-nearest neighbor (KNN) and convolutional neural network (CNN) respectively, and significantly improves localization accuracy on KNN compared to Gaussian process regression (GPR) and GPR-GAN when only a small amount of labeled fingerprints is available.

xStripeMerge: efficient wide stripe generation approach based on erasure coding storage

Meiguang ZHENG, Taofei HUA, Xinyu ZHANG, Zhigang HU

2023, 44(11):  213-224.  doi:10.11959/j.issn.1000-436x.2023217

Asbtract ( 71 )   HTML ( 7)   PDF (2833KB) ( 67 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

To address the issue of wide stripe generation in existing erasure coding storage systems, where storage scaling approaches resulted in a significant increase in wide stripe generation bandwidth and stripe merge scheme was constrained by dual-stripes, an efficient wide stripe generation approach was proposed for the wide stripe generation problem under multi stripes merging.Two key operators for the multi-stripes merging progress were defined, and the wide stripe generation problem was modeled as a combinatorial optimization problem.The efficient wide stripe generation approach xStripeMerge was proposed that prioritize the search for narrow-stripes combining schemes with small parity block transmission costs.Experimental results show that xStripeMerge can reduce the wide stripe generation bandwidth by 75% compared to the advanced storage scaling method.The time and space complexity of xStripeMerge is much better than that of the extended dual-stripes merge approach.xStripeMerge can get the wide stripe generation scheme with similar performance in a shorter period of time and it is also suitable for large-scale storage systems.

Covert communication method based on tripartite generative adversarial network

Jihong YU, Ziyan LIN, Neng YE, Kai YANG, Jianping AN

2023, 44(11):  225-236.  doi:10.11959/j.issn.1000-436x.2023215

Asbtract ( 76 )   HTML ( 23)   PDF (5951KB) ( 207 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

A novel tripartite generative adversarial network (TripartiteGAN) and a covert communication method based on TripartiteGAN were designed for jointly optimizing the transmission covertness and the demodulation accuracy of the covert message.The performance of the method was analyzed.Specifically, TripartiteGAN was used to manipulate the amplitude and phase of an input modulated covert data so that the distribution of the generated covert signal superposing the overt signal approximates to that of the overt signal for the public user.The proposed method could work with an optimum warden that needs neither to set the detection threshold manually nor to know the transmit power characteristics of the sender.Simulation results show that under the additive white Gaussian noise channel, the proposed TripartiteGAN improved the demodulation accuracy at the covert receiver end while keeping the probability of regarding the detected signal as the covert one or the overt one at the warden around 0.5.Moreover, the proposed method outperforms the existing covert communication scheme based on generative adversarial network (GAN).

Performance analysis of IRS aided downlink NOMA short-packet communication with hardware impairment

Lei YUAN, Jing XU, Mingxiu MO, Yan LEI

2023, 44(11):  237-248.  doi:10.11959/j.issn.1000-436x.2023201

Asbtract ( 70 )   HTML ( 17)   PDF (1645KB) ( 183 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

Impact of hardware impairment on the performance of intelligent reflecting surface (IRS) aided downlink two-user short-packet communication (SPC) systems using non-orthogonal multiple access over Rician channels was analyzed.Phase adjustments at the IRS were based on statistical channel state information.Firstly, the analytical expressions for average block error rates of two users were derived by using central limit theorem, moment matching method, SPC related theory, and Gaussian-Chebyshev quadrature method.Secondly, to minimize the transmission latency under the reliability constraints of two users, the optimal power allocation scheme for the considered system was given by using the proposed bisection search algorithm and one-dimensional search algorithm.Finally, simulation results verified the correctness of the theoretical analysis and showed that, compared to long-packet communication, hardware impairments have a greater impact on SPC.

PSR-SQUARES: SQL reverse synthesis system based on program space reducer

Quansheng DOU, Shun ZHANG, Hao PAN, Huixian WANG, Huanling TANG

2023, 44(11):  249-259.  doi:10.11959/j.issn.1000-436x.2023203

Asbtract ( 46 )   HTML ( 2)   PDF (3156KB) ( 106 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

In order to address the issue of rapid growth of program space in SQUARES, which led to low efficiency in program synthesis, a program space reducer based on deep neural network (DNN) was introduced into the SQUARES framework.A given ＆lt;Queried tables, Query result＆gt; pair was represented as a 2D tensor which was used as input for a DNN.And the output of the DNN was the relevance vector of the target SQL statement synthesis rules.Based on the output of the DNN, the last N rules with weak correlation to the target SQL statement were eliminated, thereby shrinking the program search space and improving the system synthesis efficiency.The architecture of DNN, the method of generating training datasets, and the training process of DNN were described in detail.Furthermore, experimental comparisons between PSR-SQUARES and other representative SQL reverse synthesis systems were conducted.The results show that the overall performance of PSR-SQUARES is superior to other synthesis systems to varying degrees, with the average synthesis time reduced from 251 s in SQUARES to 130 s and the target program synthesis success rate increased from 80% to 89%.

Comprehensive Review

Survey on adversarial attacks and defenses for object detection

Xinxin WANG, Jing CHEN, Kun HE, Zijun ZHANG, Ruiying DU, Qiao LI, Jisi SHE

2023, 44(11):  260-277.  doi:10.11959/j.issn.1000-436x.2023223

Asbtract ( 396 )   HTML ( 126)   PDF (2195KB) ( 373 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

In response to recent developments in adversarial attacks and defenses for object detection, relevant terms and concepts associated with object detection and adversarial learning were first introduced.Subsequently, according to the evolution process of the methods, a comprehensive retrospective analysis was conducted on the research achievements in the realm of adversarial attacks and defense methods for object detection.Particularly, attack methods and defense strategies were categorized based on the attacker knowledge and the deep learning lifecycle.Furthermore, an in-depth analysis and discussion of the characteristics and relationships among different approaches were provided.Lastly, considering the strengths and limitations of existing research, the imminent challenges and directions were summarized for further exploration in adversarial attack and defense of object detection.