2022 Vol. 31, No. 2

REVIEW
Backdoor Attacks on Image Classification Models in Deep Neural Networks
ZHANG Quanxin, MA Wencong, WANG Yajie, ZHANG Yaoyuan, SHI Zhiwei, LI Yuanzhang
2022, 31(2): 199-212. doi: 10.1049/cje.2021.00.126
Abstract(1352) HTML (597) PDF(283)
Abstract:
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.
SIGNAL PROCESSING
Predicting the Power Spectrum of Amplified OFDM Signals Using Higher-Order Intercept Points
YAN Siyuan, YANG Xianzhen, WANG Xiaoru, LI Fu
2022, 31(2): 213-219. doi: 10.1049/cje.2020.00.299
Abstract(330) HTML (157) PDF(71)
Abstract:
Orthogonal frequency-division multiplexing (OFDM) has been developed into a popular modulation scheme for wireless communication systems, used in applications such as LTE and 5G. In wireless communication systems, nonlinearity caused by radio frequency (RF) amplifiers will generate distortions to both passband and adjacent channels such that the transmission quality is degraded. The study of this article aims to predict the power spectrum for OFDM based signals at the output of an RF amplifier due to the nonlinearity. In this article, based on Taylor polynomial coefficients, a power spectrum expression for amplified OFDM signals in terms of intercept points (up to ${\boldsymbol{n}} $th-order) is derived. This model is useful to RF engineers in choosing and testing RF amplifiers with appropriate specifications, such as intercept points and gain, to meet the requirements of wireless standards. Measurements are carried out to confirm the results of the proposed model.
An Improved Navigation Pseudolite Signal Structure Based on the Kasami Sequences and the Pulsing Scheme
TAO Lin, SUN Junren, LI Guangchen, ZHU Bocheng
2022, 31(2): 220-226. doi: 10.1049/cje.2020.00.403
Abstract(373) HTML (157) PDF(65)
Abstract:
Pseudolites (PLs) are ground-based satellites, providing users with navigation solutions. However, implementation of the PL system leads to the near-far problem. In this paper, we proposed an improved navigation PL signal structure of combing Kasami sequences and the pulsing scheme to mitigate the near-far effect. The pulse modulation method is adopted to ensure that the PLs transmit signals at different timeslots and reduce the PL signals’ mutual interference. Additionally, we employ the small set of Kasami sequences with good cross-correlation properties to improve the anti-interference ability. A simulation test based on software is carried out to evaluate the performance of the proposed signal. The simulation proves that the improved PL signal has an impulsive power spectral density, makes it a feasible solution to mitigate the near-far effect, and performs better in the capture.
Labeled Multi-Bernoulli Maneuvering Target Tracking Algorithm via TSK Iterative Regression Model
WANG Xiaoli, XIE Weixin, LI Liangqun
2022, 31(2): 227-239. doi: 10.1049/cje.2020.00.156
Abstract(218) HTML (101) PDF(36)
Abstract:
Aiming at the problem that the existing labeled multi-Bernoulli (LMB) method has a single and fixed model set, an LMB maneuvering target tracking algorithm via Takagi-Sugeno-Kang (TSK) iterative regression multiple model is proposed. In the TSK iterative regression modeling, the feature information of the targets is analyzed and represented by multiple semantic fuzzy sets. Then the state is expanded to introduce model information, thereby the adaptive multi-model idea is incorporated into the framework of the LMB method to solve the uncertain maneuverability of moving targets. Finally, the simulation results show that the proposed algorithm can effectively achieve maneuvering target tracking in the nonlinear system.
Joint Spectrum Sensing and Spectrum Access for Defending Massive SSDF Attacks: A Novel Defense Framework
XU Zhenyu, SUN Zhiguo, GUO Lili, Muhammad Zahid Hammad, Chintha Tellambura
2022, 31(2): 240-254. doi: 10.1049/cje.2021.00.090
Abstract(245) HTML (101) PDF(52)
Abstract:
Multiple secondary users (SUs) perform collaborative spectrum sensing (CSS) in cognitive radio networks to improve the sensing performance. However, this system severely degrades with spectrum sensing data falsification (SSDF) attacks from a large number of malicious secondary users, i.e., massive SSDF attacks. To mitigate such attacks, we propose a joint spectrum sensing and spectrum access framework. During spectrum sensing, each SU compares the decisions of CSS and independent spectrum sensing (IndSS), and then the reliable decisions are adopted as its final decisions. Since the transmission slot is divided into several tiny slots, at the stage of spectrum access, each SU is assigned with a specific tiny time slot. In accordance with its independent final spectrum decisions, each node separately accesses the tiny time slot. Simulation results verify effectiveness of the proposed algorithm.
An Efficient Algebraic Solution for Moving Source Localization from Quadruple Hybrid Measurements
DING Ting, ZHAO Yongsheng, ZHAO Yongjun
2022, 31(2): 255-265. doi: 10.1049/cje.2020.00.410
Abstract(282) HTML (122) PDF(61)
Abstract:
This paper deals with the 3-D moving source localization using time difference of arrival (TDOA), frequency difference of arrival (FDOA), angle of arrival (AOA) and AOA rate measurements, gathered from a set of spatially distributed receivers. The TDOA, FDOA, AOA and AOA rate measurement equations were firstly established according to the space geometric relationship of the source relative to the receivers. Then an efficient closed-form algorithm for source position and velocity estimation from the quadruple hybrid measurements was proposed. The proposed algorithm converts the nonlinear measurement equations into a linear set of equations, which can then be used to estimate the source position and velocity applying weighted least square (WLS) minimization. In contrast to existing two-stage WLS algorithms, the proposed algorithm does not introduce any nuisance parameters and requires merely one-stage, which enables for source localization with the fewest receivers necessary. Theoretical accuracy analysis shows that the proposed algorithm reaches the Cramer-Rao lower bound, and simulation studies corroborate the efficiency and superiority of the proposed algorithm over other algorithms.
Multi-Traffic Targets Tracking Based on an Improved Structural Sparse Representation with Spatial-Temporal Constraint
YANG Honghong, SHANG Junchao, LI Jingjing, ZHANG Yumei, WU Xiaojun
2022, 31(2): 266-276. doi: 10.1049/cje.2020.00.007
Abstract(259) HTML (108) PDF(42)
Abstract:
Vehicles or pedestrians tracking is an important task in intelligent transportation system. In this paper, we propose an online multi-object tracking for intelligent traffic platform that employs improved sparse representation and structural constraint. We first build the spatial-temporal constraint via the geometric relations and appearance of tracked objects, then we construct a robust appearance model by incorporating the discriminative sparse representation with weight constraint and local sparse appearance with occlusion analysis. Finally, we complete data association by using maximum a posteriori in a Bayesian framework in the pursuit for the optimal detection estimation. Experimental results in two challenging vehicle tracking benchmark datasets show that the proposed method has a good tracking performance.
INFORMATION SECURITY AND CRYPTOLOGY
Efficient 3D Hilbert Curve Encoding and Decoding Algorithms
JIA Lianyin, LIANG Binbin, LI Mengjuan, LIU Yong, CHEN Yinong, DING Jiaman
2022, 31(2): 277-284. doi: 10.1049/cje.2020.00.171
Abstract(382) HTML (165) PDF(59)
Abstract:
Hilbert curve describes a one-to-one mapping between multidimensional space and 1D space. Most traditional 3D Hilbert encoding and decoding algorithms work on order-wise manner and are not aware of the difference between different input data and spend equivalent computing costs on them, thus resulting in a low efficiency. To solve this problem, in this paper we design efficient 3D state views for fast encoding and decoding. Based on the state views designed, a new encoding algorithm (JFK-3HE) and a new decoding algorithm (JFK-3HD) are proposed. JFK-3HE and JFK-3HD can avoid executing iteratively encoding or decoding each order by skipping the first 0s in input data, thus decreasing the complexity and improving the efficiency. Experimental results show that JFK-3HE and JFK-3HD outperform the state-of-the-arts algorithms for both uniform and skew-distributed data.
Cryptanalysis of AEGIS-128
SHI Tairong, HU Bin, GUAN Jie, WANG Senpeng
2022, 31(2): 285-292. doi: 10.1049/cje.2020.00.231
Abstract(264) HTML (113) PDF(31)
Abstract:
AEGIS, an authenticated encryption (AE) algorithm designed by H. J. Wu and B. Preneel, is one of the six winners of the Competition for Authenticated Encryption: Security, Applicability, and Robustness, which was launched by the National Institute of Standards and Technology. In this paper, we comprehensively investigate the existence of collision in the initialization of AEGIS-128 and evaluate the number of advanced encryption standard (AES) round functions involved in initialization, which reflects the resistance to differential attack. As a result, we find that there are 40 AES round functions, which is less than 50 ones claimed in the design document. We also prove that AEGIS-128 is strong enough to resist adversary who has access to partial state. In particular, we present a collision-based distinguisher and exploit it to recover the key of 4-step and 5-step (out of the full 10) AEGIS-128. The time and memory complexities are about ${{\boldsymbol{2}}}^{{\boldsymbol{29.7}}}$ and ${{\boldsymbol{2}}}^{{\boldsymbol{26}}}$ respectively. Specifically, we quantize the attack of 4-step AEGIS-128, in which we solve the technical issue of dealing with the function that does not fulfill Simon’s promise. It is noted that the nonce is not reused in our work. Although we present some results of AEGIS-128 that exceed the existed analysis, the security margin of AEGIS-128 remains large.
HiAtGang: How to Mine the Gangs Hidden Behind DDoS Attacks
ZHU Tian, QIU Xiaokang, RAO Yu, YAN Hanbing, ZHOU Yu, SHI Guixin
2022, 31(2): 293-303. doi: 10.1049/cje.2021.00.021
Abstract(127) HTML (51) PDF(14)
Abstract:
Identifying and determining behaviors of attack gangs is not only an advanced stage of the network security event tracing and analysis, but also a core step of large-scale combat and punishment of cyber attacks. Most of the work in the field of distributed denial of service (DDoS) attack analysis has focused on DDoS attack detection, and a part of the work involves the research of DDoS attack sourcing. We find that very little work has been done on the mining and analysis of DDoS attack gangs. DDoS attack gangs naturally have the attributes of human community relations. We propose a framework named HiAtGang, in which we define the concept of the gang detection in DDoS attacks and introduce the community analysis technology into DDoS attack gang analysis. Different attacker clustering algorithms are compared and analyzed. Based on analysis results of massive DDoS attack events that recorded by CNCERT/CC (The National Computer Network Emergency Response Technical Team/Coordination Center of China), the effective gang mining and attribute calibration have been achieved. More than 250 DDoS attack gangs have been successfully tracked. Our research fills the gaps in the field of the DDoS attack gang detection and has supported CNCERT/CC in publishing “Analysis Report on DDoS Attack Resources” for three consecutive years and achieved a good practical effect on combating DDoS attack crimes.
Probe Machine Based Computing Model for Maximum Clique Problem
CUI Jianzhong, YIN Zhixiang, TANG Zhen, YANG Jing
2022, 31(2): 304-312. doi: 10.1049/cje.2020.00.293
Abstract(185) HTML (76) PDF(18)
Abstract:
Probe machine (PM) is a recently reported mathematic model with massive parallelism. Herein, we presented searching the maximum clique of an undirected graph with six vertices. We constructed data library containing n sublibraries, each sublibrary corresponded to a vertex in the given graph. Then, probe library according to the induced subgraph was designed in order to search and generate all maximal cliques. Subsequently, we performed probe operation, and all maximal cliques were generated in parallel. The advantages of the proposed model lie in two aspects. On one hand, solution to NP-complete problem is generated in just one step of probe operation rather than found in vast solution space. On the other hand, the proposed model is highly parallel. The work demonstrates that PM is superior to TM in terms of searching capacity when tackling NP-complete problem.
LaTLS: A Lattice-Based TLS Proxy Protocol
ZHANG Xinglong, CHENG Qingfeng, LI Yuting
2022, 31(2): 313-321. doi: 10.1049/cje.2018.00.357
Abstract(194) HTML (89) PDF(17)
Abstract:
The function of the Internet proxy is to check and convert the data exchanged between client and server. In fact, the two-party secure communication protocol with good security is turned into an unsafe multiparty protocol. At present, there are relatively few proxy protocols that can be applied in practice. This paper analyzes the classic agent protocol mcTLS and pointed out the security issues. We focus on the security of TLS 1.3 and proposed a lattice-based multi-party proxy protocol: LaTLS. LaTLS can be proved secure in the eCK model, it can resist key-sharing attacks, counterfeiting attacks, replay attacks, and achieve forward security. Compared with traditional DH and ECDH schemes, LaTLS is more effcient. Its security is based on the shortest vector problem, therefor it has anti-quantum attack properties.
COMPUTER NETWORKS AND ARTIFICIAL INTELLIGENCE
Timely Data Delivery for Energy-Harvesting IoT Devices
LU Wenwei, GONG Siliang, ZHU Yihua
2022, 31(2): 322-336. doi: 10.1049/cje.2021.00.005
Abstract(242) HTML (105) PDF(30)
Abstract:
The devices in the Internet of things (IoT) gain capability of sustainable operation when they harvest energy from ambient sources. Fluctuation in the harvested energy may cause the energy-harvesting IoT devices to suffer from frequent energy shortage, which may bring in intolerable packet delay or packet discarding. It is important to design a low-delay packet delivery scheme that adapts to variation in the harvested energy. In this paper, we present the timely data delivery (TDD) scheme for the IoT devices. Using Markov chain, we develop a probability model for the TDD scheme, which leads to the expected number of packets delivered in an operation cycle, the expected numbers of packets waiting in the data buffer in an operation cycle and an energy-harvesting cycle, and the expected packet delay. Additionally, we formulate the optimization problem that minimizes the packet delay in the TDD scheme, and the solution to the optimization problem yields the optimal parameters for the IoT devices to determine when to harvest energy and when to deliver data under the TDD scheme. The simulation results show that the proposed TDD scheme outperforms the existing schemes in terms of packet delay.
Word-Based Method for Chinese Part-of-Speech via Parallel and Adversarial Network
HUANG Kaiyu, CAO Jingxiang, LIU Zhuang, HUANG Degen
2022, 31(2): 337-344. doi: 10.1049/cje.2020.00.411
Abstract(264) HTML (107) PDF(19)
Abstract:
Chinese part-of-speech (POS) tagging is an essential task for Chinese downstream natural language processing tasks. The accuracy of the Chinese POS task will drop dramatically by word-based methods because of the segmentation errors and the word sparsity. Also, there are several Chinese POS tagging sets with different criteria. Some of them only have a small-scale annotated corpus and are hard to train. To this end, we propose a modified word-based transformer neural network architecture. Meanwhile, we utilize an adversarial transfer learning method that splits the architecture into shared and private parts. This work directly improves the ability of the word-based model, instead of adopting a joint character-based method. Extensive experiments show that our method achieves state-of-the-art performance on all datasets, and more importantly, our method improves performance effectively for the word-based Chinese sequence labeling task.
Predicting Microbe-Disease Association Based on Heterogeneous Network and Global Graph Feature Learning
WANG Yueyue, LEI Xiujuan, PAN Yi
2022, 31(2): 345-353. doi: 10.1049/cje.2020.00.212
Abstract(447) HTML (214) PDF(73)
Abstract:
Numerous microbes inhabit human body, making a vast difference in human health. Hence, discovering associations between microbes and diseases is beneficial to disease prevention and treatment. In this study, we develop a prediction method by learning global graph feature on the heterogeneous network (called HNGFL). Firstly, a heterogeneous network is integrated by known microbe-disease associations and multiple similarities. Based on microbe Gaussian interaction profile (GIP) kernel similarity, we consider different effects of these microbes on organs in the human body to further improve microbe similarity. For disease similarity network, we combine GIP kernel similarity, disease semantic similarity and disease-symptom similarity. And then, an embedding algorithm called GraRep is used to learn global structural information for this network. According to vector feature of every node, we utilize support vector machine classifier to calculate the score for each microbe-disease pair. HNGFL achieves a reliable performance in cross validation, outperforming the compared methods. In addition, we carry out case studies of three diseases. Results show that HNGFL can be considered as a reliable method for microbe-disease association prediction.
WCM-WTrA: A Cross-Project Defect Prediction Method Based on Feature Selection and Distance-Weight Transfer Learning
LEI Tianwei, XUE Jingfeng, WANG Yong, NIU Zequn, SHI Zhiwei, ZHANG Yu
2022, 31(2): 354-366. doi: 10.1049/cje.2021.00.119
Abstract(475) HTML (228) PDF(53)
Abstract:
Cross-project defect prediction is a hot topic in the field of defect prediction. How to reduce the difference between projects and make the model have better accuracy is the core problem. This paper starts from two perspectives: feature selection and distance-weight instance transfer. We reduce the differences between projects from the perspective of feature engineering and introduce the transfer learning technology to construct a cross-project defect prediction model WCM-WTrA and multi-source model Multi-WCM-WTrA. We have tested on AEEEM and ReLink datasets, and the results show that our method has an average improvement of 23% compared with TCA+ algorithm on AEEEM datasets, and an average improvement of 5% on ReLink datasets.
Adaptive Simplified Chicken Swarm Optimization Based on Inverted S-Shaped Inertia Weight
GU Yanchun, LU Haiyan, XIANG Lei, SHEN Wanqiang
2022, 31(2): 367-386. doi: 10.1049/cje.2020.00.233
Abstract(342) HTML (149) PDF(34)
Abstract:
Considering the issues of premature convergence and low solution accuracy in solving high-dimensional problems with the basic chicken swarm optimization algorithm, an adaptive simplified chicken swarm optimization algorithm based on inverted S-shaped inertia weight (ASCSO-S) is proposed. Firstly, a simplified chicken swarm optimization algorithm is presented by removing all the chicks from the chicken swarm. Secondly, an inverted S-shaped inertia weight is designed and introduced into the updating process of the roosters and hens to dynamically adjust their moving step size and thus to improve the convergence speed and solution accuracy of the algorithm. Thirdly, in order to enhance the exploration ability of the algorithm, an adaptive updating strategy is added to the updating process of the hens. Simulation experiments on 21 classical test functions show that ASCSO-S is superior to the other comparison algorithms in terms of convergence speed, solution accuracy, and solution stability. In addition, ASCSO-S is applied to the parameter estimation of Richards model, and the test results indicate that ASCSO-S has the best fitting results compared with other three algorithms.
Knowledge Graph Completion Based on GCN of Multi-Information Fusion and High-Dimensional Structure Analysis Weight
NIU Haoran, HE Haitao, FENG Jianzhou, NIE Junlan, ZHANG Yangsen, REN Jiadong
2022, 31(2): 387-396. doi: 10.1049/cje.2021.00.080
Abstract(508) HTML (224) PDF(69)
Abstract:
Knowledge graph completion (KGC) can solve the problem of data sparsity in the knowledge graph. A large number of models for the KGC task have been proposed in recent years. However, the underutilisation of the structure information around nodes is one of the main problems of the previous KGC model, which leads to relatively single encoding information. To this end, a new KGC model that encodes and decodes the feature information is proposed. First, we adopt the subgraph sampling method to extract node structure. Moreover, the graph convolutional network (GCN) introduced the channel attention convolution encode node structure features and represent them in matrix form to fully mine the node feature information. Eventually, the high-dimensional structure analysis weight decodes the encoded matrix embeddings and then constructs the scoring function. The experimental results show that the model performs well on the datasets used.
2022-2 Contents
2022, 31(2): 397-397.
Abstract(24) HTML (12)
Abstract: