Xie Cunxiang

Dr. Xie Cunxiang


Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 August 2022 in IEEE Sensors Journal
Verified
Journal Article
Closed access
XCXie Cunxiang
Xie Cunxiang
LZLimin Zhang
Limin Zhang
ZZZhong Zhaogen
Zhong Zhaogen
Deep-learning methods are becoming increasingly popular for improving the efficiency of specific emitter identification (SEI). However, in noncooperative scenarios, it is difficult to both obtain a large number of signal samples for the training of deep-learning models and assign labels to the acquired samples. To solve these two problems, a few-shot unsupervised SEI method is proposed in this study. First, the original radio-frequency (RF) signal is preprocessed based on the Hilbert–Huang transform (HHT) to obtain the Hilbert time–frequency spectrum, which can highlight RF fingerprints (RFFs) that can be used as signal training samples. Then, the Hilbert time–frequency spectrum is input to an improved autoencoder network for training to obtain the latent vector that can represent hidden features of the received signal. Next, the clustering by fast search and find of a density peaks [i.e., density peak clustering (DPC)] algorithm is used to cluster and label the latent vector, which is later reconstructed by the improved autoencoder network to obtain training samples with labeled information. Finally, the meta-learning algorithm is used to train the few-shot SEI network under the few-shot condition, so that it can distinguish different types of RF signals, corresponding to specific emitters. The experimental results show that the unlabeled signal training samples can be clustered and labeled accurately with the proposed method to perform SEI efficiently in few-shot scenarios. In comparison with state-of-the-art solutions, the proposed method in this study exhibited its superiority with regard to few-shot unsupervised SEI.

16 May 2025 in Scientific Reports
Verified
Journal Article
Open access
PDF file
JFJun Fang
Jun Fang
XCXie Cunxiang
Xie Cunxiang
Intrusion traffic detection technology is an important network protection technology to ensure network communication security and protect users’ information privacy. To address problems relating to the low classification accuracy of current intrusion traffic detection algorithms and that most of the current research focus on closed set detection, this paper proposes a detection and classification model for open set traffic based on information maximization generative adversarial network and OpenMax algorithm. Firstly, the intrusion traffic classification model under the closed set condition is trained, and the sample activation vector is recalculated in the penultimate layer of the model by using the OpenMax algorithm. According to the activation vector of the known category, the estimated probability of the unknown category is then calculated to identify unknown traffic. Results show that the model’s classification accuracy for CICIDS2017 open set traffic in the misuse and anomaly detection experiments is above 88.5 and 88.2%, respectively. The model can effectively detect various types of unknown traffic with high detection accuracy and robustness.

13 May 2024 in IEEE Access
Verified
Journal Article
Open access
ZZZhong Zhaogen
Zhong Zhaogen
XCXie Cunxiang
Xie Cunxiang
XTXibo Tang
Xibo Tang
To solve the problem that the existing intrusion traffic detection models generally adopt machine learning algorithm and supervised deep learning algorithm, and the classification accuracy of model small samples is low, A unsupervised learning intrusion traffic classification model based on Wasserstein divergence objective for generative adversarial nets (WGAN-div) and information maximizing generative adversarial nets (Info GAN) is presented. The algorithm uses generative adversarial network to optimize the sampling of unbalanced data sets and effectively improves the feature extraction capability of small samples of the model. Firstly, the unbalanced data training set is oversampled by WGAN-div to improve the data distribution. Then, the non-data part is processed by independent thermal coding and integrated with the data part to reduce the complexity of pretreatment. Finally, the Info GAN model is used for data training. Performance evaluation and algorithm performance comparison were carried out in NSL-KDD, CICIDS2017 and UNSW-NB15 data sets. The experimental results show that the accuracy of multi-classification task is 91.1%, 97.1%, 79.9% respectively, and the accuracy of binary classification task is 90.9%, 96.9%, 86.1% respectively. Compared with the classical deep learning algorithm, the Info GAN model has higher accuracy and lower false positive rate, and has higher reliability and engineering application value.

15 December 2022 in IEEE Access
Verified
Journal Article
Open access
XCXie Cunxiang
Xie Cunxiang
LZLimin Zhang
Limin Zhang
ZZZhong Zhaogen
Zhong Zhaogen
Obtaining larger category-label-containing training signal datasets in non-cooperative scenarios is difficult. Moreover, employing smaller labeled signal datasets for specific emitter identification is technically challenging. Therefore, we propose a novel method for few-shot SEI. We first design a bispectral analysis and Radon transformation-based signal preprocessing scheme to obtain feature vectors that effectively characterize the radio frequency fingerprints. The feature vectors are then fed to a network model for feature learning. Moreover, a meta-learning algorithm is applied to the network model to adapt to few-shot feature learning. The conventional meta-learning algorithm is improved to develop a novel algorithm involving latent embedding optimization for meta-learning. The proposed method extracts low-dimensional key features from high-dimensional input data and evaluates the distance and degree of feature dispersion. The resulting information is employed in sample point prediction. The algorithm effectively achieves few-shot SEI and offers stable and efficient recognition after training with a minimum of forty samples. This method identifies emitter individuals under multiple modulation types and exhibits scalability in identifying the emitter numbers. Moreover, it offers adaptability in identifying the emitter individuals under multiple propagation channel types.

25 July 2023 in Electronics
Verified
Journal Article
Open access
PDF file
XCXie Cunxiang
Xie Cunxiang
LZLimin Zhang
Limin Zhang
ZZZhong Zhaogen
Zhong Zhaogen
This study focused on the construction of a spatiotemporal knowledge graph for ship activities. First, a ship activity ontology model was proposed to describe the entities and relations of ship activities. Then, maritime event text data were utilized as the ship activity dataset, where entities and relations were extracted to form triplets. Thus, the data layer was populated, completing the construction of the ship activity spatiotemporal knowledge graph. The process of extracting triplets involved initially inputting the text sentences into the Bidirectional Encoder Representations from Transformers (BERT) model for pretraining to obtain vector representations of characters. These representations were then fed into a lattice long short-term memory network (Lattice-LSTM) for further processing. The resulting hidden vectors h1,h2,⋯,hn were input into the conditional random field (CRF) to perform named entity recognition. The recognized entities were then labeled in the original sentences and input into another BERT-Lattice-LSTM network. The resulting hidden vectors h′1,h′2,⋯,h′n were fed into a relation classifier, which output the relation between the two labeled entities, completing the extraction of entity–relation triplets. In experiments, the proposed method achieved triplet extraction performance exceeding 90% for three different evaluation metrics: Precision, Recall, and F1-measure.

6 May 2023 in Applied Sciences
Verified
Journal Article
Open access
PDF file
XCXie Cunxiang
Xie Cunxiang
LZLimin Zhang
Limin Zhang
ZZZhong Zhaogen
Zhong Zhaogen
Entity alignment helps discover and link entities from different knowledge graphs (KGs) that refer to the same real-world entity, making it a critical technique for KG fusion. Most entity alignment methods are based on knowledge representation learning, which uses a mapping function to project entities from different KGs into a unified vector space and align them based on calculated similarities. However, this process requires sufficient pre-aligned entity pairs. To address this problem, this study proposes an entity alignment method based on joint learning of entity and attribute representations. Structural embeddings are learned using the triples modeling method based on TransE and PTransE and extracted from the embedding vector space utilizing semantic information from direct and multi-step relation paths. Simultaneously, attribute character embeddings are learned using the N-gram-based compositional function to encode a character sequence for the attribute values, followed by TransE to model attribute triples in the embedding vector space to obtain attribute character embedding vectors. By learning the structural and attribute character embeddings simultaneously, the structural embeddings of entities from different KGs can be transferred into a unified vector space. Lastly, the similarities in the structural embedding of different entities were calculated to perform entity alignment. The experimental results showed that the proposed method performed well on the DBP15K and DWK100K datasets, and it outperformed currently available entity alignment methods by 16.8, 27.5, and 24.0% in precision, recall, and F1 measure, respectively.