Changqing An
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
Changqing An has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Aichun Huang
Nankai University, China
1 shared publication - Chao Zhang
Tsinghua University, Beijing, China and Zhongguancun Laboratory, Beijing, China
1 shared publication - Chengwei Dai
XekRung Team, Alibaba Group, China
1 shared publication - Fang Meng
Network Research Center, Tsinghua University, Beijing, China. phone number: +86-10-62795818-6926
1 shared publication - Fasheng Miao
Tsinghua University, Beijing, China
1 shared publication Fengyuan Ma
Beijing Univ Posts & Telecommun, Inst Network Technol, Beijing 100876, Peoples R China
1 shared publication
Latest Publications
- Open Access
- 22 October 2026
- Published by MDPI AG in Preprints.org
Large Language Models for Cybersecurity: A Survey of Data, Training, Models, Agent Systems, Evaluation, and Applications
Advances in language understanding, code reasoning, and tool use are enabling cybersecurity agents to progress from question answering and single-step analysis to vulnerability research, incident investigation, and multistage task execution. These tasks require models to gather information, adapt their strategies, and verify results through sustained interaction, making capability development increasingly dependent on data, training, and runtime mechanisms. This survey examines research on Large Language Models (LLMs) for cybersecurity across six dimensions: data, training, models, agent systems, evaluation, and applications. We first examine how security corpora, instruction examples, interaction trajectories, and preference data are constructed, and how executable environments support training and validation. We then compare domain-adaptive continued pretraining, supervised fine-tuning, preference optimization, agentic reinforcement learning, and distillation; analyze the capabilities of security-specialized and general-purpose models; and review mechanisms for planning, tool use, memory management, coordination, and verification. Drawing on representative benchmarks and experimental results, we assess progress and identify major bottlenecks across applications in threat intelligence, vulnerability analysis and remediation, malware analysis, security operations, and digital forensics. Based on this systematic review and analysis, we argue that cybersecurity LLMs are progressing from acquiring domain knowledge to learning how to perform interactive tasks. High-quality security data supports domain learning and task-specific supervision. Executable environments extend this training to multistep analysis and tool use, with outcome feedback guiding strategy refinement. Task quality and cost depend on both the model and its harness: the model reasons and selects actions, while the harness organizes information, tools, and execution state. Improvements in vulnerability discovery and threat analysis must also support effective remediation and timely response to yield lasting security benefits. Future research should develop agentic training and model–harness co-optimization methods that use validated attack–defense interactions to support continual improvement. Learning from operational tasks could further extend model capabilities and their contribution to cybersecurity practice.
- 22 October 2026
- Published by MDPI AG in Preprints.org
Open AccessLarge Language Models for Cybersecurity: A Survey of Data, Training, Models, Agent Systems, Evaluation, and Applications
Advances in language understanding, code reasoning, and tool use are enabling cybersecurity agents to progress from question answering and single-step analysis to vulnerability research, incident investigation, and multistage task execution. These tasks require models to gather information, adapt their strategies, and verify results through sustained interaction, making capability development increasingly dependent on data, training, and runtime mechanisms. This survey examines research on Large Language Models (LLMs) for cybersecurity across six dimensions: data, training, models, agent systems, evaluation, and applications. We first examine how security corpora, instruction examples, interaction trajectories, and preference data are constructed, and how executable environments support training and validation. We then compare domain-adaptive continued pretraining, supervised fine-tuning, preference optimization, agentic reinforcement learning, and distillation; analyze the capabilities of security-specialized and general-purpose models; and review mechanisms for planning, tool use, memory management, coordination, and verification. Drawing on representative benchmarks and experimental results, we assess progress and identify major bottlenecks across applications in threat intelligence, vulnerability analysis and remediation, malware analysis, security operations, and digital forensics. Based on this systematic review and analysis, we argue that cybersecurity LLMs are progressing from acquiring domain knowledge to learning how to perform interactive tasks. High-quality security data supports domain learning and task-specific supervision. Executable environments extend this training to multistep analysis and tool use, with outcome feedback guiding strategy refinement. Task quality and cost depend on both the model and its harness: the model reasons and selects actions, while the harness organizes information, tools, and execution state. Improvements in vulnerability discovery and threat analysis must also support effective remediation and timely response to yield lasting security benefits. Future research should develop agentic training and model–harness co-optimization methods that use validated attack–defense interactions to support continual improvement. Learning from operational tasks could further extend model capabilities and their contribution to cybersecurity practice.
- 1 February 2020
- Published by Tsinghua University Press in Tsinghua Science and Technology
- Vol. 25 (1) , 140-148
MD-AVB: A multi-manifold based available bandwidth prediction algorithm
- 1 February 2020
- Published by Tsinghua University Press in Tsinghua Science and Technology
- Vol. 25 (1) , 140-148
MD-AVB: A multi-manifold based available bandwidth prediction algorithm
- 1 November 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Network Traffic Capture and Storage for IPv6 Network
Network traffic characteristics in operational network is crucial to network design and operation. As the network bandwidth grows, it's difficult for capturing and storing a detailed record for every packet. In this paper we present a traffic capture system which is implemented at network processor. Both lab experiment and real network operation prove that the capture system is able to process gigabit line rate with a good stability. A data compression method focusing on IPv6 address redundancy and combining flow and packet information together is given for IPv6 trace and alleviates the data storage pressure. Tests show it can compress the storage several fold without losing detailed trace information.
- 1 November 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Network Traffic Capture and Storage for IPv6 Network
Network traffic characteristics in operational network is crucial to network design and operation. As the network bandwidth grows, it's difficult for capturing and storing a detailed record for every packet. In this paper we present a traffic capture system which is implemented at network processor. Both lab experiment and real network operation prove that the capture system is able to process gigabit line rate with a good stability. A data compression method focusing on IPv6 address redundancy and combining flow and packet information together is given for IPv6 trace and alleviates the data storage pressure. Tests show it can compress the storage several fold without losing detailed trace information.
- 1 June 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 5 (15503607) , 2177-2182
Scalable Double Filter Structure for Port Scan Detection
Port scan detection is very important to predict network intrusions and prevent viruses from spreading. Many networks deploy Network Intrusion Detection Systems (NIDS) to detect port scans in real-time. However, most NIDS are perflow based. They are not scalable on high speed links since it is infeasible to maintain the states of numerous flows. In this paper, we propose a scalable scheme for real-time port scan detection without keeping any per-flow state. We use a double-filter structure to find out pairs which connect to more than N pairs in T time. The experimental results on real network traces show that our scheme can find out those over-threshold pairs with high accuracy. It is easy to scale our scheme to high speed environments due to its little memory consumption and fast processing pipeline.
- 1 June 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 5 (15503607) , 2177-2182
Scalable Double Filter Structure for Port Scan Detection
Port scan detection is very important to predict network intrusions and prevent viruses from spreading. Many networks deploy Network Intrusion Detection Systems (NIDS) to detect port scans in real-time. However, most NIDS are perflow based. They are not scalable on high speed links since it is infeasible to maintain the states of numerous flows. In this paper, we propose a scalable scheme for real-time port scan detection without keeping any per-flow state. We use a double-filter structure to find out pairs which connect to more than N pairs in T time. The experimental results on real network traces show that our scheme can find out those over-threshold pairs with high accuracy. It is easy to scale our scheme to high speed environments due to its little memory consumption and fast processing pipeline.