Hong Qin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hong Qin has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Debiao He
Wuhan University
3 shared publications - Qi Feng
Wuhan University
3 shared publications - Min Luo
School of Cyber Science and Engineering, Wuhan University, Wuhan, China
2 shared publications - Guihe Gu
School of Computer Science, Wuhan University, Wuhan 430072, China
1 shared publication - Hao Wang
Shandong Normal University
1 shared publication - Huawei Li
School of Artificial Intelligence, Wuhan University, Wuhan, China
1 shared publication
Latest Publications
- Open Access
Q-GrAM: Fine-Grained Image–Text Retrieval via Grouped Query Routing and Conditional Query Modulation
Existing image–text retrieval methods often compute cross-modal similarity using global single-vector representations. Although efficient for coarse semantic alignment, such compressed representations are limited when textual queries involve fine-grained semantics, including objects, attributes, relations, and their compositional structures. This paper focuses on fine-grained text-to-image retrieval and proposes Q-GrAM, a retrieval-oriented adaptation of the BLIP-2 Q-Former. Instead of treating Q-Former queries as a homogeneous set, Q-GrAM partitions a fixed query budget into semantically differentiated groups. A text-guided router assigns token-level semantic demands to query groups, while query conditional initialization modulates each group according to group-level textual summaries. The resulting grouped visual query features are matched with text tokens through a group-aware late interaction scorer, and auxiliary routing balance and inter-group diversity regularization are introduced to stabilize semantic specialization. Experiments on MS-COCO 5K, Flickr30K, and Flickr30K-CFQ show that Q-GrAM achieves strong text-to-image retrieval performance against both global embedding baselines and representative fine-grained image–text matching methods, while maintaining competitive bidirectional retrieval performance. These results demonstrate the effectiveness of structured, text-conditioned Q-Former query specialization for fine-grained text-driven image search.
Open AccessQ-GrAM: Fine-Grained Image–Text Retrieval via Grouped Query Routing and Conditional Query Modulation
Existing image–text retrieval methods often compute cross-modal similarity using global single-vector representations. Although efficient for coarse semantic alignment, such compressed representations are limited when textual queries involve fine-grained semantics, including objects, attributes, relations, and their compositional structures. This paper focuses on fine-grained text-to-image retrieval and proposes Q-GrAM, a retrieval-oriented adaptation of the BLIP-2 Q-Former. Instead of treating Q-Former queries as a homogeneous set, Q-GrAM partitions a fixed query budget into semantically differentiated groups. A text-guided router assigns token-level semantic demands to query groups, while query conditional initialization modulates each group according to group-level textual summaries. The resulting grouped visual query features are matched with text tokens through a group-aware late interaction scorer, and auxiliary routing balance and inter-group diversity regularization are introduced to stabilize semantic specialization. Experiments on MS-COCO 5K, Flickr30K, and Flickr30K-CFQ show that Q-GrAM achieves strong text-to-image retrieval performance against both global embedding baselines and representative fine-grained image–text matching methods, while maintaining competitive bidirectional retrieval performance. These results demonstrate the effectiveness of structured, text-conditioned Q-Former query specialization for fine-grained text-driven image search.
- 11 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. PP (16) , 1
RDFDS: A Federated Learning Defense Framework in IOV with Fast Distillation Synthesis and Clustering-Based Aggregation
Federated Learning (FL) has emerged as a promising paradigm to address data silos and privacy concerns in artificial intelligence applications. It shows great potential in the Internet of Vehicles (IoV), where sensitive vehicle data must remain local. However, FL is highly vulnerable to model poisoning attacks, especially under non-IID distributions and high malicious participation. Existing defenses, most relying on gradient similarity across clients, often fail under such heterogeneous settings. To address this, we propose RDFDS, a robust defense framework for FL in IoV. RDFDS consists of two tightly coupled modules: (1) a fast distillation-based synthesis module that accelerates knowledge distillation via a lightweight preprocessing step, suitable for real-time vehicular environments; (2) a Gas-KMeans clustering aggregation module that adaptively assigns aggregation weights by identifying the benign majority through Gap Statistics, preserving reliable contributions without requiring global or cross-client reference data. Extensive experiments on MNIST, CIFAR-10, and GTSRB demonstrate that RDFDS consistently outperforms existing defenses in accuracy and robustness, particularly under extreme non-IID conditions and high malicious participation, highlighting its practicality for real-world IoV scenarios.
- 11 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. PP (16) , 1
RDFDS: A Federated Learning Defense Framework in IOV with Fast Distillation Synthesis and Clustering-Based Aggregation
Federated Learning (FL) has emerged as a promising paradigm to address data silos and privacy concerns in artificial intelligence applications. It shows great potential in the Internet of Vehicles (IoV), where sensitive vehicle data must remain local. However, FL is highly vulnerable to model poisoning attacks, especially under non-IID distributions and high malicious participation. Existing defenses, most relying on gradient similarity across clients, often fail under such heterogeneous settings. To address this, we propose RDFDS, a robust defense framework for FL in IoV. RDFDS consists of two tightly coupled modules: (1) a fast distillation-based synthesis module that accelerates knowledge distillation via a lightweight preprocessing step, suitable for real-time vehicular environments; (2) a Gas-KMeans clustering aggregation module that adaptively assigns aggregation weights by identifying the benign majority through Gap Statistics, preserving reliable contributions without requiring global or cross-client reference data. Extensive experiments on MNIST, CIFAR-10, and GTSRB demonstrate that RDFDS consistently outperforms existing defenses in accuracy and robustness, particularly under extreme non-IID conditions and high malicious participation, highlighting its practicality for real-world IoV scenarios.
- Open Access
- 20 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
- Vol. 26 (2) , 1-22
Secure and Dropout-Resilient Three-Party Clustering Based on Cloud-Edge-Client Collaboration
Clustering algorithms, as the core technology in data analysis, can extract potential patterns and regularities from complex data. However, deploying k -means clustering on resource-limited devices remains a challenge. Despite the promise of cloud computing, outsourcing data to a remote cloud leads to high latency and privacy risks. Moreover, the stability and speed of cloud can be affected by the state of network and configuration, which leads to computation error. Therefore, we design a secure and dropout-resilient k -means clustering scheme based on cloud-edge-client collaboration architecture. In our scheme, cloud server simply generates multiplication triples in pre-processing phase and can be offline. In online phase, IoT devices secretly share the raw sensing data with three edge servers. Then edge servers accomplish the clustering task interactively. We propose four basic protocols based on vector space secret sharing, including Euclidean distance, comparison, minimum and division protocols. By applying these protocols, we construct a clustering scheme that can tolerate the exit of one edge server and corruption of two edge servers. Since edge servers are generally located in trusted environment, we allow them to reconstruct clustering result and provide low-latency and high-reliability service. We prove that the basic protocols and clustering scheme are secure against semi-honest adversary. We conduct the experiments on two realistic datasets, showing that our scheme has good efficiency and is suitable for practical application.
- 20 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
- Vol. 26 (2) , 1-22
Open AccessSecure and Dropout-Resilient Three-Party Clustering Based on Cloud-Edge-Client Collaboration
Clustering algorithms, as the core technology in data analysis, can extract potential patterns and regularities from complex data. However, deploying k -means clustering on resource-limited devices remains a challenge. Despite the promise of cloud computing, outsourcing data to a remote cloud leads to high latency and privacy risks. Moreover, the stability and speed of cloud can be affected by the state of network and configuration, which leads to computation error. Therefore, we design a secure and dropout-resilient k -means clustering scheme based on cloud-edge-client collaboration architecture. In our scheme, cloud server simply generates multiplication triples in pre-processing phase and can be offline. In online phase, IoT devices secretly share the raw sensing data with three edge servers. Then edge servers accomplish the clustering task interactively. We propose four basic protocols based on vector space secret sharing, including Euclidean distance, comparison, minimum and division protocols. By applying these protocols, we construct a clustering scheme that can tolerate the exit of one edge server and corruption of two edge servers. Since edge servers are generally located in trusted environment, we allow them to reconstruct clustering result and provide low-latency and high-reliability service. We prove that the basic protocols and clustering scheme are secure against semi-honest adversary. We conduct the experiments on two realistic datasets, showing that our scheme has good efficiency and is suitable for practical application.
- 25 October 2024
- Published by Springer Nature
- p. 310-328
Outsourced and Robust Multi-party Computation with Identifying Malicious Behavior and Application to Machine Learning
- 25 October 2024
- Published by Springer Nature
- p. 310-328
Outsourced and Robust Multi-party Computation with Identifying Malicious Behavior and Application to Machine Learning
- 3 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 36 (5) , 1919-1934
Cryptographic Primitives in Privacy-Preserving Machine Learning: A Survey
Advances in machine learning have enabled a broad range of complex applications, such as image recognition, recommendation system and machine translation. Data plays an important role in our increasingly complex and diverse environments, and this also reinforces the importance of data privacy in machine learning-enabled applications. Although there are a number of literature survey articles on machine learning, only a few studies have investigated the cryptographic primitives used in privacy-preserving machine learning (PPML). In other words, there is no, or limited, systematization of knowledge (SoK) that provides a comprehensive introduction to cryptography that have been deployed in PPML. In this paper, we firstly introduce some basic concepts such as machine learning tasks and processes. Then, we review and systematize the cryptographic primitives used in PPML. We analyze these existing privacy-preserving schemes in their learning process, especially training and inference. Finally, we conclude our survey and provide an outlook on future trends and research directions in the field.
- 3 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 36 (5) , 1919-1934
Cryptographic Primitives in Privacy-Preserving Machine Learning: A Survey
Advances in machine learning have enabled a broad range of complex applications, such as image recognition, recommendation system and machine translation. Data plays an important role in our increasingly complex and diverse environments, and this also reinforces the importance of data privacy in machine learning-enabled applications. Although there are a number of literature survey articles on machine learning, only a few studies have investigated the cryptographic primitives used in privacy-preserving machine learning (PPML). In other words, there is no, or limited, systematization of knowledge (SoK) that provides a comprehensive introduction to cryptography that have been deployed in PPML. In this paper, we firstly introduce some basic concepts such as machine learning tasks and processes. Then, we review and systematize the cryptographic primitives used in PPML. We analyze these existing privacy-preserving schemes in their learning process, especially training and inference. Finally, we conclude our survey and provide an outlook on future trends and research directions in the field.