Dongyang Zhan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Dongyang Zhan has a total of 67 co-authors. Here are the top co-authors by number of shared publications:
- Lin Ye
School of Cyberspace Science, Harbin Institute of Technology, Harbin, China
14 shared publications - Hongli Zhang
School of Cyberspace Science, Harbin Institute of Technology, Harbin, China
12 shared publications Lin Ye
10 shared publicationsHongli Zhang
9 shared publicationsXiangzhan Yu
7 shared publicationsBinxing Fang
6 shared publications
Latest Publications
- Open Access
- 7 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 3552-3556
AgentSysFilter: A Multi-agent Analysis Approach for Scenario-Aware Syscall Limitation
- 7 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 3552-3556
Open AccessAgentSysFilter: A Multi-agent Analysis Approach for Scenario-Aware Syscall Limitation
- Open Access
- 6 August 2026
- Published by MDPI AG in Preprints.org
A Survey on the Unique Security of Autonomous and Collaborative LLM Agents: Threats, Defenses, and Futures
The rapid evolution of Large Language Models (LLMs) from static text generators to autonomous agents has revolutionized their ability to perceive, reason, and act within complex environments. However, this transition shifts the security problem from model-level robustness to system-level engineering. The conflation of instructions and data, persistent cognitive states, and untrusted coordination create structural risks that are not adequately captured by traditional adversarial robustness. To address the fragmented nature of the existing literature, this article presents a comprehensive and systematic survey of the security landscape for LLM-based agents. We propose a novel, structure-aware taxonomy that categorizes threats into three distinct paradigms: (1) External Interaction Attacks, which exploit vulnerabilities in perception interfaces and tool usage; (2) Internal Cognitive Attacks, which compromise the integrity of reasoning chains and memory mechanisms; and (3) Multi-Agent Collaboration Attacks, which manipulate communication protocols and collective decision-making. Adapting to this threat landscape, we systematize existing mitigation strategies into a unified defense framework that includes input sanitization, cognitive fortification, and collaborative consensus. In addition, we provide an in-depth comparative analysis of agent-specific security evaluation benchmarks. The survey concludes by outlining critical open problems and future research directions, aiming to foster the development of next-generation agents that are not only autonomous but also secure and trustworthy. The companion repository is available at https://github.com/sunyinggang/LLM-Agent-Security-Survey.
- 6 August 2026
- Published by MDPI AG in Preprints.org
Open AccessA Survey on the Unique Security of Autonomous and Collaborative LLM Agents: Threats, Defenses, and Futures
The rapid evolution of Large Language Models (LLMs) from static text generators to autonomous agents has revolutionized their ability to perceive, reason, and act within complex environments. However, this transition shifts the security problem from model-level robustness to system-level engineering. The conflation of instructions and data, persistent cognitive states, and untrusted coordination create structural risks that are not adequately captured by traditional adversarial robustness. To address the fragmented nature of the existing literature, this article presents a comprehensive and systematic survey of the security landscape for LLM-based agents. We propose a novel, structure-aware taxonomy that categorizes threats into three distinct paradigms: (1) External Interaction Attacks, which exploit vulnerabilities in perception interfaces and tool usage; (2) Internal Cognitive Attacks, which compromise the integrity of reasoning chains and memory mechanisms; and (3) Multi-Agent Collaboration Attacks, which manipulate communication protocols and collective decision-making. Adapting to this threat landscape, we systematize existing mitigation strategies into a unified defense framework that includes input sanitization, cognitive fortification, and collaborative consensus. In addition, we provide an in-depth comparative analysis of agent-specific security evaluation benchmarks. The survey concludes by outlining critical open problems and future research directions, aiming to foster the development of next-generation agents that are not only autonomous but also secure and trustworthy. The companion repository is available at https://github.com/sunyinggang/LLM-Agent-Security-Survey.
- Open Access
- 5 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 675-686
DA-MLAD: Drift-Decomposed Meta-Learning for Continual Log Anomaly Detection in Supercomputing Systems
- 5 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 675-686
Open AccessDA-MLAD: Drift-Decomposed Meta-Learning for Continual Log Anomaly Detection in Supercomputing Systems
- Open Access
A High-performance Real-time Container File Monitoring Approach Based on Virtual Machine Introspection
As cloud computing continues to advance and become an integral part of modern IT infrastructure, container security has emerged as a critical factor in ensuring the smooth operation of cloud-native applications. An attacker can attack the service in the container or even perform the container escape attack by tampering with the files. Monitoring container files is important for APT detection and cyberspace security. Existing file monitoring methods are usually based on host operating system or virtual machine introspection to protect file security in real time. The methods based on the host operating system usually monitor file operations in the host operating system. However, when the container escapes to the host, the host operating system will no longer be secure, so these methods face the problem of weak security. Aiming at the problems of low security and high overload introduced in existing container file monitoring, a high-performance container file monitoring method based on virtual machine introspection is proposed. The experimental results show that the proposed approach can effectively monitor the container files and introduce an acceptable monitoring overload.
Open AccessA High-performance Real-time Container File Monitoring Approach Based on Virtual Machine Introspection
As cloud computing continues to advance and become an integral part of modern IT infrastructure, container security has emerged as a critical factor in ensuring the smooth operation of cloud-native applications. An attacker can attack the service in the container or even perform the container escape attack by tampering with the files. Monitoring container files is important for APT detection and cyberspace security. Existing file monitoring methods are usually based on host operating system or virtual machine introspection to protect file security in real time. The methods based on the host operating system usually monitor file operations in the host operating system. However, when the container escapes to the host, the host operating system will no longer be secure, so these methods face the problem of weak security. Aiming at the problems of low security and high overload introduced in existing container file monitoring, a high-performance container file monitoring method based on virtual machine introspection is proposed. The experimental results show that the proposed approach can effectively monitor the container files and introduce an acceptable monitoring overload.
- Open Access
A Practical Adversarial Attack against Sequence-based Deep Learning Malware Classifiers
Sequence-based deep learning models (e.g., RNNs), can detect malware by analyzing its behavioral sequences. Meanwhile, these models are susceptible to adversarial attacks. Attackers can create adversarial samples that alter the sequence characteristics of behavior sequences to deceive malware classifiers. The existing methods for generating adversarial samples typically involve deleting or replacing crucial behaviors in the original data sequences, or inserting benign behaviors that may violate the behavior constraints. However, these methods that directly manipulate sequences make adversarial samples difficult to implement or apply in practice. In this paper, we propose an adversarial attack approach based on Deep Q-Network and a heuristic backtracking search strategy, which can generate perturbation sequences that satisfy practical conditions for successful attacks. Subsequently, we utilize a novel transformation approach that maps modifications back to the source code, thereby avoiding the need to directly modify the behavior log sequences. We conduct an evaluation of our approach, and the results confirm its effectiveness in generating adversarial samples from real-world malware behavior sequences, which have a high success rate in evading anomaly detection models. Furthermore, our approach is practical and can generate adversarial samples while maintaining the functionality of the modified software.
Open AccessA Practical Adversarial Attack against Sequence-based Deep Learning Malware Classifiers
Sequence-based deep learning models (e.g., RNNs), can detect malware by analyzing its behavioral sequences. Meanwhile, these models are susceptible to adversarial attacks. Attackers can create adversarial samples that alter the sequence characteristics of behavior sequences to deceive malware classifiers. The existing methods for generating adversarial samples typically involve deleting or replacing crucial behaviors in the original data sequences, or inserting benign behaviors that may violate the behavior constraints. However, these methods that directly manipulate sequences make adversarial samples difficult to implement or apply in practice. In this paper, we propose an adversarial attack approach based on Deep Q-Network and a heuristic backtracking search strategy, which can generate perturbation sequences that satisfy practical conditions for successful attacks. Subsequently, we utilize a novel transformation approach that maps modifications back to the source code, thereby avoiding the need to directly modify the behavior log sequences. We conduct an evaluation of our approach, and the results confirm its effectiveness in generating adversarial samples from real-world malware behavior sequences, which have a high success rate in evading anomaly detection models. Furthermore, our approach is practical and can generate adversarial samples while maintaining the functionality of the modified software.
- Open Access
Anomaly Detection in Industrial Control Systems Based on Cross-Domain Representation Learning
Industrial control systems (ICSs) are widely used in industry, and their security and stability are very important. Once the ICS is attacked, it may cause serious damage. Therefore, it is very important to detect anomalies in ICSs. ICS can monitor and manage physical devices remotely using communication networks. The existing anomaly detection approaches mainly focus on analyzing the security of network traffic or sensor data. However, the behaviors of different domains (e.g., network traffic and sensor physical status) of ICSs are correlated, so it is difficult to comprehensively identify anomalies by analyzing only a single domain. In this paper, an anomaly detection approach based on cross-domain representation learning in ICSs is proposed, which can learn the joint features of multi-domain behaviors and detect anomalies within different domains. After constructing a cross-domain graph that can represent the behaviors of multiple domains in ICSs, our approach can learn the joint features of them by leveraging graph neural networks. Since anomalies behave differently in different domains, we leverage a multi-task learning approach to identify anomalies in different domains separately and perform joint training. The experimental results show that the performance of our approach is better than existing approaches for identifying anomalies in ICSs.
Open AccessAnomaly Detection in Industrial Control Systems Based on Cross-Domain Representation Learning
Industrial control systems (ICSs) are widely used in industry, and their security and stability are very important. Once the ICS is attacked, it may cause serious damage. Therefore, it is very important to detect anomalies in ICSs. ICS can monitor and manage physical devices remotely using communication networks. The existing anomaly detection approaches mainly focus on analyzing the security of network traffic or sensor data. However, the behaviors of different domains (e.g., network traffic and sensor physical status) of ICSs are correlated, so it is difficult to comprehensively identify anomalies by analyzing only a single domain. In this paper, an anomaly detection approach based on cross-domain representation learning in ICSs is proposed, which can learn the joint features of multi-domain behaviors and detect anomalies within different domains. After constructing a cross-domain graph that can represent the behaviors of multiple domains in ICSs, our approach can learn the joint features of them by leveraging graph neural networks. Since anomalies behave differently in different domains, we leverage a multi-task learning approach to identify anomalies in different domains separately and perform joint training. The experimental results show that the performance of our approach is better than existing approaches for identifying anomalies in ICSs.
- Open Access
Exploring and Exploiting the Resource Isolation Attack Surface of WebAssembly Containers
Recently, the WebAssembly (or Wasm) technology has been rapidly evolving, with many runtimes actively under development, providing cross-platform secure sandboxes for Wasm modules to run as portable containers. Compared with Docker, which isolates applications at the operating system level, Wasm runtimes provide more security mechanisms, such as linear memory, type checking, and protected call stacks. Although Wasm is designed with security in mind and considered to be a more secure container runtime, various security challenges have arisen, and researchers have focused on the security of Wasm runtimes, such as discovering vulnerabilities or proposing new security mechanisms to achieve robust isolation. However, we have observed that the resource isolation is not well protected by the current Wasm runtimes, and attackers can exhaust the host's resources to interfere with the execution of other container instances by exploiting the WASI/WASIX interfaces. And the attack surface has not been well explored and measured. In this paper, we explore the resource isolation attack surface of Wasm runtimes systematically by proposing several static Wasm runtime analysis approaches. Based on the analysis results, we propose several exploitation strategies to break the resource isolation of Wasm runtimes. The experimental results show that malicious Wasm instances can not only consume large amounts of system resources on their own but also introduce high workloads into other components of the underlying operating system, leading to a substantial performance degradation of the whole system. In addition, the mitigation approaches have also been discussed.
Open AccessExploring and Exploiting the Resource Isolation Attack Surface of WebAssembly Containers
Recently, the WebAssembly (or Wasm) technology has been rapidly evolving, with many runtimes actively under development, providing cross-platform secure sandboxes for Wasm modules to run as portable containers. Compared with Docker, which isolates applications at the operating system level, Wasm runtimes provide more security mechanisms, such as linear memory, type checking, and protected call stacks. Although Wasm is designed with security in mind and considered to be a more secure container runtime, various security challenges have arisen, and researchers have focused on the security of Wasm runtimes, such as discovering vulnerabilities or proposing new security mechanisms to achieve robust isolation. However, we have observed that the resource isolation is not well protected by the current Wasm runtimes, and attackers can exhaust the host's resources to interfere with the execution of other container instances by exploiting the WASI/WASIX interfaces. And the attack surface has not been well explored and measured. In this paper, we explore the resource isolation attack surface of Wasm runtimes systematically by proposing several static Wasm runtime analysis approaches. Based on the analysis results, we propose several exploitation strategies to break the resource isolation of Wasm runtimes. The experimental results show that malicious Wasm instances can not only consume large amounts of system resources on their own but also introduce high workloads into other components of the underlying operating system, leading to a substantial performance degradation of the whole system. In addition, the mitigation approaches have also been discussed.
- 30 August 2025
- Published by Elsevier in Information Fusion
- Vol. 126
Distributed machine learning for next-generation communication networks: A survey on privacy, fairness, efficiency, and trade-offs
- 30 August 2025
- Published by Elsevier in Information Fusion
- Vol. 126
Distributed machine learning for next-generation communication networks: A survey on privacy, fairness, efficiency, and trade-offs
- 19 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 22 (3) , 2505-2518
Anomaly Detection in Industrial Control Systems Based on Cross-Domain Representation Learning
Industrial control systems (ICSs) are widely used in industry, and their security and stability are very important. Once the ICS is attacked, it may cause serious damage. Therefore, it is very important to detect anomalies in ICSs. ICS can monitor and manage physical devices remotely using communication networks. The existing anomaly detection approaches mainly focus on analyzing the security of network traffic or sensor data. However, the behaviors of different domains (e.g., network traffic and sensor physical status) of ICSs are correlated, so it is difficult to comprehensively identify anomalies by analyzing only a single domain. In this article, an anomaly detection approach based on cross-domain representation learning in ICSs is proposed, which can learn the joint features of multi-domain behaviors and detect anomalies within different domains. After constructing a cross-domain graph that can represent the behaviors of multiple domains in ICSs, our approach can learn the joint features of them by leveraging graph neural networks. Since anomalies behave differently in different domains, we leverage a multi-task learning approach to identify anomalies in different domains separately and perform joint training. The experimental results show that the performance of our approach is better than existing approaches for identifying anomalies in ICSs.
- 19 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 22 (3) , 2505-2518
Anomaly Detection in Industrial Control Systems Based on Cross-Domain Representation Learning
Industrial control systems (ICSs) are widely used in industry, and their security and stability are very important. Once the ICS is attacked, it may cause serious damage. Therefore, it is very important to detect anomalies in ICSs. ICS can monitor and manage physical devices remotely using communication networks. The existing anomaly detection approaches mainly focus on analyzing the security of network traffic or sensor data. However, the behaviors of different domains (e.g., network traffic and sensor physical status) of ICSs are correlated, so it is difficult to comprehensively identify anomalies by analyzing only a single domain. In this article, an anomaly detection approach based on cross-domain representation learning in ICSs is proposed, which can learn the joint features of multi-domain behaviors and detect anomalies within different domains. After constructing a cross-domain graph that can represent the behaviors of multiple domains in ICSs, our approach can learn the joint features of them by leveraging graph neural networks. Since anomalies behave differently in different domains, we leverage a multi-task learning approach to identify anomalies in different domains separately and perform joint training. The experimental results show that the performance of our approach is better than existing approaches for identifying anomalies in ICSs.