Tao Xue
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Tao Xue has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
Haibin Zhang
School of Cyber Engineering, Xidian University, Xian, China
3 shared publicationsYanbin Wang
College of Computer Science and Technology, Shenzhen MSU-BIT University, Hangzhou, China
2 shared publications- Bo Luo
The University of Kansas, Lawrence, KS, USA
1 shared publication - Boyang Zhang
Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China
1 shared publication - Dan Meng
Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China
1 shared publication - Gang Li
Deakin University
1 shared publication
Latest Publications
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (8) , 16148-16168
NLACP Dataset Expansion With AI: Leveraging ChatGPT for Policy Generation
Natural language access control policies (NLACPs) play a pivotal role in access control systems, yet the scarcity of high-quality public datasets severely hinders advancements in natural language processing (NLP)/machine learning (ML)-driven policy automation. Existing datasets, constrained by labor-intensive manual curation, confidentiality barriers, and rigid synthesis methods, suffer from limited scale, semantic inconsistency, and poor adaptability to emerging domains such as smart healthcare and the Internet of Vehicles (IoV). To address this critical gap, we propose GenPy, a novel framework leveraging large language models (LLMs) like ChatGPT and domain-specific knowledge graphs to systematically generate scalable, semantically rich NLACP datasets. GenPy pioneers the integration of LLMs into NLACP synthesis, offering three core innovations: a generation module combining knowledge graphs with prompt engineering to produce diverse, domain-aligned policies; an annotation module employing few-shot learning to filter nonpolicy statements and ensure quality; and a policy management module resolving redundancies and conflicts while enabling conflict injection for optimization research. By embedding domain semantics (e.g., IoV compliance with automotive data laws), GenPy enriches traditional datasets and supports fine-grained policy requirements. Experiments demonstrate that GenPy generates more than 5000 high-quality NLACPs across multiple domains, achieving an $F1$ -score of exceeding 93% in NLACP classification tasks. Notably, augmenting the classification models in RAGent with GenPy-generated data improves NLACP classification accuracy by 2%–9%, validating its practical utility. Our work establishes the first LLM-based pipeline that integrates knowledge graphs with customizable prompts to generate semantically rich NLACPs, supporting fine-grained access control models, NLACP identification, and policy management across diverse security requirements. We provide an open-source repository of policies, code, and technical frameworks to catalyze future research in access control automation. The dataset and implementation are publicly available at https://github.com/fe1w0/NLACP-LLM
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (8) , 16148-16168
NLACP Dataset Expansion With AI: Leveraging ChatGPT for Policy Generation
Natural language access control policies (NLACPs) play a pivotal role in access control systems, yet the scarcity of high-quality public datasets severely hinders advancements in natural language processing (NLP)/machine learning (ML)-driven policy automation. Existing datasets, constrained by labor-intensive manual curation, confidentiality barriers, and rigid synthesis methods, suffer from limited scale, semantic inconsistency, and poor adaptability to emerging domains such as smart healthcare and the Internet of Vehicles (IoV). To address this critical gap, we propose GenPy, a novel framework leveraging large language models (LLMs) like ChatGPT and domain-specific knowledge graphs to systematically generate scalable, semantically rich NLACP datasets. GenPy pioneers the integration of LLMs into NLACP synthesis, offering three core innovations: a generation module combining knowledge graphs with prompt engineering to produce diverse, domain-aligned policies; an annotation module employing few-shot learning to filter nonpolicy statements and ensure quality; and a policy management module resolving redundancies and conflicts while enabling conflict injection for optimization research. By embedding domain semantics (e.g., IoV compliance with automotive data laws), GenPy enriches traditional datasets and supports fine-grained policy requirements. Experiments demonstrate that GenPy generates more than 5000 high-quality NLACPs across multiple domains, achieving an $F1$ -score of exceeding 93% in NLACP classification tasks. Notably, augmenting the classification models in RAGent with GenPy-generated data improves NLACP classification accuracy by 2%–9%, validating its practical utility. Our work establishes the first LLM-based pipeline that integrates knowledge graphs with customizable prompts to generate semantically rich NLACPs, supporting fine-grained access control models, NLACP identification, and policy management across diverse security requirements. We provide an open-source repository of policies, code, and technical frameworks to catalyze future research in access control automation. The dataset and implementation are publicly available at https://github.com/fe1w0/NLACP-LLM
- Open Access
- 5 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
EfficientSTNet: A Deep Learning Approach for Multi-Class DDoS Detection
DDoS attacks are highly destructive, capable of targeting multiple devices simultaneously and posing significant threats to network systems. Therefore, it is increasingly essential to develop effective and reliable detection methods to ensure network security. Although deep learning techniques improve DDoS detection by autonomously learning feature representations, they still face challenges due to the high dimensionality and noise inherent in network traffic data, which impair computational efficiency and real-time responsiveness. Moreover, their limited adaptability to multi-class classification hinders accurate differentiation of specific attack types. This paper proposes a multi-class DDoS attack detection method based on multi-scale feature modeling—EfficientSTNet. The method employs a selective deep autoencoder for feature selection, then uses a convolutional neural network (CNN) to capture hidden spatial features in network data. Subsequently, a self-attention mechanism extracts temporal features and captures contextual dependencies among different features, improving DDoS attack detection accuracy. Finally, a fully connected layer and normalization compute the probability distribution across different classes, enabling multi-class DDoS detection. Additionally, the method integrates convolution decomposition techniques and residual network structures to accelerate model convergence and improve inference speed. The proposed EfficientSTNet model can identify different types of DDoS attacks, achieving an overall classification accuracy of 99.14% and F1-scores of 97.97% or higher across all categories, demonstrating strong practical application potential.
- 5 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Open AccessEfficientSTNet: A Deep Learning Approach for Multi-Class DDoS Detection
DDoS attacks are highly destructive, capable of targeting multiple devices simultaneously and posing significant threats to network systems. Therefore, it is increasingly essential to develop effective and reliable detection methods to ensure network security. Although deep learning techniques improve DDoS detection by autonomously learning feature representations, they still face challenges due to the high dimensionality and noise inherent in network traffic data, which impair computational efficiency and real-time responsiveness. Moreover, their limited adaptability to multi-class classification hinders accurate differentiation of specific attack types. This paper proposes a multi-class DDoS attack detection method based on multi-scale feature modeling—EfficientSTNet. The method employs a selective deep autoencoder for feature selection, then uses a convolutional neural network (CNN) to capture hidden spatial features in network data. Subsequently, a self-attention mechanism extracts temporal features and captures contextual dependencies among different features, improving DDoS attack detection accuracy. Finally, a fully connected layer and normalization compute the probability distribution across different classes, enabling multi-class DDoS detection. Additionally, the method integrates convolution decomposition techniques and residual network structures to accelerate model convergence and improve inference speed. The proposed EfficientSTNet model can identify different types of DDoS attacks, achieving an overall classification accuracy of 99.14% and F1-scores of 97.97% or higher across all categories, demonstrating strong practical application potential.
- 15 July 2025
- Published by Elsevier in Computer Networks
- Vol. 270
Edge computing for IoT: Novel insights from a comparative analysis of access control models
- 15 July 2025
- Published by Elsevier in Computer Networks
- Vol. 270
Edge computing for IoT: Novel insights from a comparative analysis of access control models
- 8 February 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (2) , 1104-1123
SparkAC: Fine-Grained Access Control in Spark for Secure Data Sharing and Analytics
With the development of computing and communication technologies, an extremely large amount of data has been collected, stored, utilized, and shared, while new security and privacy challenges arise. Existing access control mechanisms provided by big data platforms have limitations in granularity and expressiveness. In this paper, we present SparkAC, a novel access control mechanism for secure data sharing and analysis in Spark. In particular, we first propose a purpose-aware access control (PAAC) model, which introduces new concepts of data processing purpose and data operation purpose and an automatic purpose analysis algorithm that identifies purposes from data analytics operations and queries. Moreover, we develop a unified access control mechanism that implements PAAC model in two modules. GuardSpark++ supports structured data access control in Spark Catalyst and GuardDAG supports unstructured data access control in Spark core. Finally, we evaluate GuardSpark++ and GuardDAG with multiple data sources, applications, and data analytics engines. Experimental results show that SparkAC provides effective access control functionalities with very small (GuardSpark++) or medium (GuardDAG) performance overhead
- 8 February 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (2) , 1104-1123
SparkAC: Fine-Grained Access Control in Spark for Secure Data Sharing and Analytics
With the development of computing and communication technologies, an extremely large amount of data has been collected, stored, utilized, and shared, while new security and privacy challenges arise. Existing access control mechanisms provided by big data platforms have limitations in granularity and expressiveness. In this paper, we present SparkAC, a novel access control mechanism for secure data sharing and analysis in Spark. In particular, we first propose a purpose-aware access control (PAAC) model, which introduces new concepts of data processing purpose and data operation purpose and an automatic purpose analysis algorithm that identifies purposes from data analytics operations and queries. Moreover, we develop a unified access control mechanism that implements PAAC model in two modules. GuardSpark++ supports structured data access control in Spark Catalyst and GuardDAG supports unstructured data access control in Spark core. Finally, we evaluate GuardSpark++ and GuardDAG with multiple data sources, applications, and data analytics engines. Experimental results show that SparkAC provides effective access control functionalities with very small (GuardSpark++) or medium (GuardDAG) performance overhead