Hao Hu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Hao Hu has a total of 64 co-authors. Here are the top co-authors by number of shared publications:
Hongqi Zhang
State Key Laboratory of Mathematical Engineering and Advanced Computing, Zhengzhou, Henan, China
9 shared publications- Yuling Liu
Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China
7 shared publications - Jinglei Tan
State Key Laboratory of Mathematical Engineering and Advanced Computing, Information Engineering University, Zhengzhou, China
5 shared publications Yuling Liu
5 shared publications- Hengwei Zhang
Information Engineering University, Zhengzhou, China
4 shared publications Bin Yu
Zhengzhou Information Science and Technology Institute
3 shared publications
Latest Publications
- Open Access
- 14 October 2026
- Published by Springer Nature in Cybersecurity
- Vol. 9 (1) , 240
MIRAGE: manipulation-immune RAG with adaptive guard against textual evidence attacks
Retrieval-augmented generation (RAG) substantially improves large language models (LLMs) on knowledge-intensive tasks by incorporating external knowledge, yet open knowledge sources introduce prominent security risks. The SafeRAG benchmark identifies four novel textual evidence manipulation attacks, namely silver noise, inter-context conflict, soft advertisement, and White Denial-of-Service (White DoS), which bypass conventional defense components and threaten both the security and availability of RAG systems. Existing single-stage defenses fail to capture the implicit malicious characteristics of these attacks, while multi-stage frameworks are limited by fixed parameters, weak conflict resolution, and insufficient detection of implicit toxicity and availability threats. To address these limitations, we propose MIRAGE, a two-stage adaptive defense framework combining retrieval-side purification and generation-side resolution. The first stage applies attack-characteristic-aware dynamic K-means clustering and dynamic ROUGE-L threshold filtering to eliminate malicious documents. The second stage employs attack-adaptive knowledge integration rules alongside an enhanced self-evaluation mechanism that jointly verifies factual accuracy, implicit toxicity, and availability. Under the static, non-adaptive attack setting defined by SafeRAG, experiments on DeepSeek-v3.2 show that MIRAGE achieves average improvements of 20.9% in Attack Failure Rate (AFR), 11.5% in Retrieval Accuracy (RA), and 14.1% in F1(avg) over baselines across all four attack scenarios, providing effective technical support for the stable deployment of RAG systems in complex adversarial environments.
- 14 October 2026
- Published by Springer Nature in Cybersecurity
- Vol. 9 (1) , 240
Open AccessMIRAGE: manipulation-immune RAG with adaptive guard against textual evidence attacks
Retrieval-augmented generation (RAG) substantially improves large language models (LLMs) on knowledge-intensive tasks by incorporating external knowledge, yet open knowledge sources introduce prominent security risks. The SafeRAG benchmark identifies four novel textual evidence manipulation attacks, namely silver noise, inter-context conflict, soft advertisement, and White Denial-of-Service (White DoS), which bypass conventional defense components and threaten both the security and availability of RAG systems. Existing single-stage defenses fail to capture the implicit malicious characteristics of these attacks, while multi-stage frameworks are limited by fixed parameters, weak conflict resolution, and insufficient detection of implicit toxicity and availability threats. To address these limitations, we propose MIRAGE, a two-stage adaptive defense framework combining retrieval-side purification and generation-side resolution. The first stage applies attack-characteristic-aware dynamic K-means clustering and dynamic ROUGE-L threshold filtering to eliminate malicious documents. The second stage employs attack-adaptive knowledge integration rules alongside an enhanced self-evaluation mechanism that jointly verifies factual accuracy, implicit toxicity, and availability. Under the static, non-adaptive attack setting defined by SafeRAG, experiments on DeepSeek-v3.2 show that MIRAGE achieves average improvements of 20.9% in Attack Failure Rate (AFR), 11.5% in Retrieval Accuracy (RA), and 14.1% in F1(avg) over baselines across all four attack scenarios, providing effective technical support for the stable deployment of RAG systems in complex adversarial environments.
- Open Access
- 7 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1534-1546
Making Theft Useless: Adulteration-Based Protection of Proprietary Knowledge Graphs in GraphRAG Systems
- 7 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1534-1546
Open AccessMaking Theft Useless: Adulteration-Based Protection of Proprietary Knowledge Graphs in GraphRAG Systems
- 18 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (4) , 9549-9566
Enhancing Cloud Network Resilience via a Robust LLM-Empowered Multi-Agent Reinforcement Learning Framework
While virtualization and resource pooling empower cloud networks with structural flexibility and elastic scalability, they inevitably expand the attack surface and challenge cyber resilience. Reinforcement Learning (RL)-based defense strategies have been developed to optimize resource deployment and isolation policies under adversarial conditions, aiming to enhance system resilience by maintaining and restoring network availability. However, existing approaches lack robustness as they require retraining to adapt to dynamic changes in network structure, node scale, attack strategies, and attack intensity. Furthermore, the lack of Human-in-the-Loop (HITL) support limits interpretability and flexibility. To address these limitations, we propose CyberOps-Bots, a hierarchical multi agent reinforcement learning framework empowered by Large Language Models (LLMs). Inspired by MITRE ATT&CK's “Tactics-Techniques” model, CyberOps-Bots features a two-layer architecture: (1) An upper-level LLM agent with four mod ules—ReAct planning, IPDRR-based perception, long-short term memory, and action/tool integration—performs global awareness, human intent recognition, and tactical planning; (2) Lower-level RL agents, developed via heterogeneous separated pre-training, execute atomic defense actions within localized network regions. This synergy preserves LLM adaptability and interpretability while ensuring reliable RL execution. Experiments on real cloud datasets show that, compared to state-of-the-art algorithms, CyberOps-Bots maintains network availability 68.5% higher and achieves a 34.7% jumpstart performance gain when shifting the scenarios without retraining. To our knowledge, this is the first study to establish a robust LLM-RL framework with HITL support for cloud defense.
- 18 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (4) , 9549-9566
Enhancing Cloud Network Resilience via a Robust LLM-Empowered Multi-Agent Reinforcement Learning Framework
While virtualization and resource pooling empower cloud networks with structural flexibility and elastic scalability, they inevitably expand the attack surface and challenge cyber resilience. Reinforcement Learning (RL)-based defense strategies have been developed to optimize resource deployment and isolation policies under adversarial conditions, aiming to enhance system resilience by maintaining and restoring network availability. However, existing approaches lack robustness as they require retraining to adapt to dynamic changes in network structure, node scale, attack strategies, and attack intensity. Furthermore, the lack of Human-in-the-Loop (HITL) support limits interpretability and flexibility. To address these limitations, we propose CyberOps-Bots, a hierarchical multi agent reinforcement learning framework empowered by Large Language Models (LLMs). Inspired by MITRE ATT&CK's “Tactics-Techniques” model, CyberOps-Bots features a two-layer architecture: (1) An upper-level LLM agent with four mod ules—ReAct planning, IPDRR-based perception, long-short term memory, and action/tool integration—performs global awareness, human intent recognition, and tactical planning; (2) Lower-level RL agents, developed via heterogeneous separated pre-training, execute atomic defense actions within localized network regions. This synergy preserves LLM adaptability and interpretability while ensuring reliable RL execution. Experiments on real cloud datasets show that, compared to state-of-the-art algorithms, CyberOps-Bots maintains network availability 68.5% higher and achieves a 34.7% jumpstart performance gain when shifting the scenarios without retraining. To our knowledge, this is the first study to establish a robust LLM-RL framework with HITL support for cloud defense.
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (14) , 31294-31307
Attack Path Planning in 5G-ICPS Penetration Testing: Leveraging TGNN and DRL for Large-Scale Network
Path planning constitutes a critical component of penetration testing for 5G-ICPS networks. The diversity of interfaces and protocols necessitates deep analysis of vulnerability exploitation methods and cross-protocol combination strategies, significantly increasing attack path planning complexity. Furthermore, dynamic network slice configurations and physical-information coupling effects drive continuous topological evolution, causing state-space explosion and challenging path planning under uncertainty with incomplete information. To address these issues, we construct a temporal attack graph modeling 5G-ICPS attack processes, and design a GraphSAGE-based environment representation encoder. This encoder undergoes multi-tiered self-supervised pre-training, employing node-level and graph-level training to encode diverse reinforcement learning environments across attack scenarios into fixed-dimensional vector representations. This achieves decoupling from underlying topology, vulnerability specifics, and security configurations, effectively mitigating state-space explosion in large-scale networks. Subsequently, we cluster highly similar vulnerabilities and filter invalid attack actions using three typical 5G-ICPS attack constraints, compressing the agent’s exploration space. Especially, we design a customized reward function that dynamically incentivizes/penalizes actions based on compromised assets. Experimental results demonstrate significant improvements: penetration testing invalid action rates decrease from 22.3% to 7.5%, while average steps to achieve attack targets reduce by >54%. These advancements effectively reduce penetration testing costs and increase attack success rates.
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (14) , 31294-31307
Attack Path Planning in 5G-ICPS Penetration Testing: Leveraging TGNN and DRL for Large-Scale Network
Path planning constitutes a critical component of penetration testing for 5G-ICPS networks. The diversity of interfaces and protocols necessitates deep analysis of vulnerability exploitation methods and cross-protocol combination strategies, significantly increasing attack path planning complexity. Furthermore, dynamic network slice configurations and physical-information coupling effects drive continuous topological evolution, causing state-space explosion and challenging path planning under uncertainty with incomplete information. To address these issues, we construct a temporal attack graph modeling 5G-ICPS attack processes, and design a GraphSAGE-based environment representation encoder. This encoder undergoes multi-tiered self-supervised pre-training, employing node-level and graph-level training to encode diverse reinforcement learning environments across attack scenarios into fixed-dimensional vector representations. This achieves decoupling from underlying topology, vulnerability specifics, and security configurations, effectively mitigating state-space explosion in large-scale networks. Subsequently, we cluster highly similar vulnerabilities and filter invalid attack actions using three typical 5G-ICPS attack constraints, compressing the agent’s exploration space. Especially, we design a customized reward function that dynamically incentivizes/penalizes actions based on compromised assets. Experimental results demonstrate significant improvements: penetration testing invalid action rates decrease from 22.3% to 7.5%, while average steps to achieve attack targets reduce by >54%. These advancements effectively reduce penetration testing costs and increase attack success rates.
- Open Access
- 16 December 2025
- Published by Springer Nature in Cybersecurity
- Vol. 8 (1) , 116
A zero-shot self-improving NER method for cyber threat intelligence via knowledge injection
The rapid evolution of cyber threats demands efficient entity extraction from Cyber Threat Intelligence (CTI) reports to support proactive analysis and sharing. Current methods for CTI extraction falter due to a lack of domain knowledge, which can lead to the overlooking of critical entities. Moreover, the hallucinations in LLM’s outputs result in insufficient accuracy. To address these limitations, we propose a zero-shot, self-improving NER method for CTI via knowledge injection. The framework consists of four modules: a domain knowledge extractor, a reliable data annotator, a high-consistency annotation filter, and a self-retrieval reasoner. The domain knowledge extractor enhances LLM comprehension of specialized threat intelligence, while the others work in a multi-stage reasoning process to mitigate hallucinations by generating, filtering, and reasoning upon high-consistency data. These modules collaborate to improve the model’s entity recognition ability through continuous in-context learning. Experimental results show that under strict zero-shot conditions, the proposed method achieves F1 scores of 67.7%, 61.41%, 74.56%, and 65.83% on the LLM-TIKG, APT-NER, LADDER, and CDTier datasets, respectively. This represents an improvement of 7.66% over the average F1 score of other baseline methods, demonstrating superior adaptability in low-resource security scenarios.
- 16 December 2025
- Published by Springer Nature in Cybersecurity
- Vol. 8 (1) , 116
Open AccessA zero-shot self-improving NER method for cyber threat intelligence via knowledge injection
The rapid evolution of cyber threats demands efficient entity extraction from Cyber Threat Intelligence (CTI) reports to support proactive analysis and sharing. Current methods for CTI extraction falter due to a lack of domain knowledge, which can lead to the overlooking of critical entities. Moreover, the hallucinations in LLM’s outputs result in insufficient accuracy. To address these limitations, we propose a zero-shot, self-improving NER method for CTI via knowledge injection. The framework consists of four modules: a domain knowledge extractor, a reliable data annotator, a high-consistency annotation filter, and a self-retrieval reasoner. The domain knowledge extractor enhances LLM comprehension of specialized threat intelligence, while the others work in a multi-stage reasoning process to mitigate hallucinations by generating, filtering, and reasoning upon high-consistency data. These modules collaborate to improve the model’s entity recognition ability through continuous in-context learning. Experimental results show that under strict zero-shot conditions, the proposed method achieves F1 scores of 67.7%, 61.41%, 74.56%, and 65.83% on the LLM-TIKG, APT-NER, LADDER, and CDTier datasets, respectively. This represents an improvement of 7.66% over the average F1 score of other baseline methods, demonstrating superior adaptability in low-resource security scenarios.
- 4 October 2025
- Published by Elsevier in Computer Networks
- Vol. 273
LLM4Game: Multi-agent reinforcement learning with knowledge injection for dynamic defense resource allocation in cloud storage
- 4 October 2025
- Published by Elsevier in Computer Networks
- Vol. 273
LLM4Game: Multi-agent reinforcement learning with knowledge injection for dynamic defense resource allocation in cloud storage
- Open Access
- 6 July 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (13) , 7915
Knowledge-Graph- and GCN-Based Domain Chinese Long Text Classification Method
In order to solve the current problems in domain long text classification tasks, namely, the long length of a document, which makes it difficult for the model to capture key information, and the lack of expert domain knowledge, which leads to insufficient classification accuracy, a domain long text classification model based on a knowledge graph and a graph convolutional neural network is proposed. BERT is used to encode the text, and each word’s corresponding vector is used as a node for the graph convolutional neural network so that the initialized vector contains rich semantic information. Using the trained entity–relationship extraction model, the entity-to-entity–relationships in the document are extracted and used as the edges of the graph convolutional neural network, together with syntactic dependency information. The graph structure mask is used to learn about edge relationships and edge types to further enhance the learning ability of the model for semantic dependencies between words. The method further improves the accuracy of domain long text classification by fusing knowledge features and data features. Experiments on three long text classification datasets—IFLYTEK, THUCNews, and the Chinese corpus of Fudan University—show accuracy improvements of 8.8%, 3.6%, and 2.6%, respectively, relative to the BERT model.
- 6 July 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (13) , 7915
Open AccessKnowledge-Graph- and GCN-Based Domain Chinese Long Text Classification Method
In order to solve the current problems in domain long text classification tasks, namely, the long length of a document, which makes it difficult for the model to capture key information, and the lack of expert domain knowledge, which leads to insufficient classification accuracy, a domain long text classification model based on a knowledge graph and a graph convolutional neural network is proposed. BERT is used to encode the text, and each word’s corresponding vector is used as a node for the graph convolutional neural network so that the initialized vector contains rich semantic information. Using the trained entity–relationship extraction model, the entity-to-entity–relationships in the document are extracted and used as the edges of the graph convolutional neural network, together with syntactic dependency information. The graph structure mask is used to learn about edge relationships and edge types to further enhance the learning ability of the model for semantic dependencies between words. The method further improves the accuracy of domain long text classification by fusing knowledge features and data features. Experiments on three long text classification datasets—IFLYTEK, THUCNews, and the Chinese corpus of Fudan University—show accuracy improvements of 8.8%, 3.6%, and 2.6%, respectively, relative to the BERT model.
- Open Access
- 27 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (6) , 4719-4732
WF-MTD: Evolutionary Decision Method for Moving Target Defense Based on Wright-Fisher Process
The limitations of the professional knowledge and cognitive capabilities of both attackers and defenders mean that moving target attack-defense conflicts are not completely rational, which makes it difficult to select optimal moving target defense strategies difficult for use in real-world attack-defense scenarios. Starting from the imperfect rationality of both attack-defense, we construct a Wright-Fisher process-based moving target defense strategy evolution model called WF-MTD. In our method, we introduce rationality parameters to describe the strategy learning capabilities of both the attacker and the defender. By solving for the evolutionarily stable equilibrium, we develop a method for selecting the optimal defense strategy for moving targets and describe the evolution trajectories of the attack-defense strategies. Our experimental results in our example of a typical network information system show that WF-MTD selects appropriate MTD strategies in different states along different attack paths, with good effectiveness and broad applicability. In addition, compared with no hopping strategy, fixed periodic route hopping strategy, and random periodic route hopping strategy, the route hopping strategy based on WF-MTD increase defense payoffs by 58.7%, 27.6%, and 24.6%, respectively.
- 27 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (6) , 4719-4732
Open AccessWF-MTD: Evolutionary Decision Method for Moving Target Defense Based on Wright-Fisher Process
The limitations of the professional knowledge and cognitive capabilities of both attackers and defenders mean that moving target attack-defense conflicts are not completely rational, which makes it difficult to select optimal moving target defense strategies difficult for use in real-world attack-defense scenarios. Starting from the imperfect rationality of both attack-defense, we construct a Wright-Fisher process-based moving target defense strategy evolution model called WF-MTD. In our method, we introduce rationality parameters to describe the strategy learning capabilities of both the attacker and the defender. By solving for the evolutionarily stable equilibrium, we develop a method for selecting the optimal defense strategy for moving targets and describe the evolution trajectories of the attack-defense strategies. Our experimental results in our example of a typical network information system show that WF-MTD selects appropriate MTD strategies in different states along different attack paths, with good effectiveness and broad applicability. In addition, compared with no hopping strategy, fixed periodic route hopping strategy, and random periodic route hopping strategy, the route hopping strategy based on WF-MTD increase defense payoffs by 58.7%, 27.6%, and 24.6%, respectively.
- Open Access
- 21 August 2021
- Published by Wiley in Security and Communication Networks
- Vol. 2021, 1-16
Optimal Network Defense Strategy Selection Method: A Stochastic Differential Game Model
In a real-world network confrontation process, attack and defense actions change rapidly and continuously. The network environment is complex and dynamically random. Therefore, attack and defense strategies are inevitably subject to random disturbances during their execution, and the transition of the network security state is affected accordingly. In this paper, we construct a network security state transition model by referring to the epidemic evolution process, use Gaussian noise to describe random effects during the strategy execution, and introduce a random disturbance intensity factor to describe the degree of random effects. On this basis, we establish an attack-defense stochastic differential game model, propose a saddle point equilibrium solution method, and provide an algorithm to select the optimal defense strategy. Our method achieves real-time defense decision-making in network attack-defense scenarios with random disturbances and has better real-time performance and practicality than current methods. Results of a simulation experiment show that our model and algorithm are effective and feasible.
- 21 August 2021
- Published by Wiley in Security and Communication Networks
- Vol. 2021, 1-16
Open AccessOptimal Network Defense Strategy Selection Method: A Stochastic Differential Game Model
In a real-world network confrontation process, attack and defense actions change rapidly and continuously. The network environment is complex and dynamically random. Therefore, attack and defense strategies are inevitably subject to random disturbances during their execution, and the transition of the network security state is affected accordingly. In this paper, we construct a network security state transition model by referring to the epidemic evolution process, use Gaussian noise to describe random effects during the strategy execution, and introduce a random disturbance intensity factor to describe the degree of random effects. On this basis, we establish an attack-defense stochastic differential game model, propose a saddle point equilibrium solution method, and provide an algorithm to select the optimal defense strategy. Our method achieves real-time defense decision-making in network attack-defense scenarios with random disturbances and has better real-time performance and practicality than current methods. Results of a simulation experiment show that our model and algorithm are effective and feasible.
- 31 October 2020
- Published by Korean Society for Internet Information (KSII) in KSII Transactions on Internet and Information Systems
- Vol. 14 (10) , 4157-4175
SOCMTD: Selecting Optimal Countermeasure for Moving Target Defense Using Dynamic Game
KSII Transactions on Internet and Information Systems TIIS Paper Details
- 31 October 2020
- Published by Korean Society for Internet Information (KSII) in KSII Transactions on Internet and Information Systems
- Vol. 14 (10) , 4157-4175
SOCMTD: Selecting Optimal Countermeasure for Moving Target Defense Using Dynamic Game
KSII Transactions on Internet and Information Systems TIIS Paper Details