Shenghao Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Shenghao Liu has a total of 98 co-authors. Here are the top co-authors by number of shared publications:
- Xianjun Deng
School of Cyber Science and Engineering, Huazhong University of Science and Technology, Wuhan, China
35 shared publications - Laurence Tianruo Yang
School of Computer Science and Technology, Hainan University, Haikou, China and Department of Computer Science, St. Francis Xavier University, Antigonish, Nova Scotia, Canada
14 shared publications - Xiaoxuan Fan
Hubei Key Laboratory of Distributed System Security, Hubei Engineering Research Center on Big Data Security, School of Cyber Science and Engineering, Huazhong University of Science and Technology, Wuhan, China
14 shared publications - Chenlu Zhu
Hubei Key Laboratory of Distributed System Security, Hubei Engineering Research Center on Big Data Security, School of Cyber Science and Engineering, Huazhong University of Science and Technology, Wuhan, China
9 shared publications - Bang Wang
Huazhong University of Science and Technology, Wuhan, China
8 shared publications - Jong Hyuk Park
Seoul National University of Science and Technology
8 shared publications
Latest Publications
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (7) , 11032-11046
Heterogeneous Hypergraph Enhanced Trust Recommendation in Mobile Social Networks
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (7) , 11032-11046
Heterogeneous Hypergraph Enhanced Trust Recommendation in Mobile Social Networks
- 6 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP, 1-15
PrivGNN: Privacy-Preserving Graph Neural Network Inference Based on Collaborative Computing for Cyber-Physical-Social Systems
Graph neural networks (GNNs) are gaining significant attention for enabling “cognitive intelligence” by processing complex, non-Euclidean data, particularly in cyber-physical-social systems (CPSS). These systems rely on real-time decision-making to optimize interactions across physical, social, and digital environments. Collaborative learning further enhances GNNs by allowing distributed entities to learn together from shared data, improving scalability and intelligence. However, the intricate structure of graph data and the increasing use of GNNs raise serious privacy concerns, particularly when sensitive data from distributed artificial intelligence of things (AIoT) devices of CPSS is involved. Although several privacy-preserving GNN schemes have been proposed, they often incur prohibitive communication and computation costs. To address the challenges, we propose PrivGNN, an efficient and privacy-preserving system for GNN inference with the help of multiple cloud servers. This system leverages novel variants of secret sharing techniques to develop secure multiparty computing protocols that safeguard data privacy. Our security analysis and performance evaluations confirm that PrivGNN ensures robust security, high efficiency, and scalability in cloud-based inference services. Almost each subprotocol reduces protocol communication costs from 33% to 40% and computation times from 25% to 30% in the LAN without compromising security. Benchmarking on the Cora and Pubmed datasets shows that PrivGNN cuts overall communication expenses by at least 30%, highlighting its practical value for privacy-sensitive, large-scale applications in AIoT and CPSS.
- 6 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP, 1-15
PrivGNN: Privacy-Preserving Graph Neural Network Inference Based on Collaborative Computing for Cyber-Physical-Social Systems
Graph neural networks (GNNs) are gaining significant attention for enabling “cognitive intelligence” by processing complex, non-Euclidean data, particularly in cyber-physical-social systems (CPSS). These systems rely on real-time decision-making to optimize interactions across physical, social, and digital environments. Collaborative learning further enhances GNNs by allowing distributed entities to learn together from shared data, improving scalability and intelligence. However, the intricate structure of graph data and the increasing use of GNNs raise serious privacy concerns, particularly when sensitive data from distributed artificial intelligence of things (AIoT) devices of CPSS is involved. Although several privacy-preserving GNN schemes have been proposed, they often incur prohibitive communication and computation costs. To address the challenges, we propose PrivGNN, an efficient and privacy-preserving system for GNN inference with the help of multiple cloud servers. This system leverages novel variants of secret sharing techniques to develop secure multiparty computing protocols that safeguard data privacy. Our security analysis and performance evaluations confirm that PrivGNN ensures robust security, high efficiency, and scalability in cloud-based inference services. Almost each subprotocol reduces protocol communication costs from 33% to 40% and computation times from 25% to 30% in the LAN without compromising security. Benchmarking on the Cora and Pubmed datasets shows that PrivGNN cuts overall communication expenses by at least 30%, highlighting its practical value for privacy-sensitive, large-scale applications in AIoT and CPSS.
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Sensor Networks
- Vol. 22 (3) , 1-24
MVSTA: Multi-View Spatio-Temporal Correlation Awareness Network for Traffic Data Imputation
Road traffic data imputation is an essential component of Intelligent Transportation Systems (ITS). However, the spatio-temporal characteristics in traffic data are complex and diverse, and existing methods are unable to comprehensively extract them. This article, through in-depth observations of traffic system dynamics, innovatively taxonomizes the complex spatio-temporal dependencies into four key dimensions: Geographical Spatial Correlations (GSC), Latent Spatial Correlations (LSC), Intra-sensor Temporal Correlations (ITC), and Cross-sensor Temporal Correlations (CTC). Motivated by this taxonomy, we aim at constructing a novel framework that holistically utilizes these spatio-temporal correlations to improve imputation accuracy. Subsequently, we propose a Multi-View Spatio-Temporal Correlation Awareness Network (MVSTA) for traffic data imputation, which incorporates two specifically designed modules: a Unified Spatial Correlation Awareness module (USCA) and a Collaborative Temporal Correlation Awareness module (CTCA). The USCA integrates GSC and LSC into a unified representation by jointly modeling physical proximity and data-driven dependencies. The CTCA collaboratively extracts ITC and CTC by capturing both local temporal patterns in individual sensors and interactive dynamics across different sensors. Extensive experiments on three real-world traffic datasets demonstrate that MVSTA significantly outperforms all baselines, validating the effectiveness of our proposed taxonomy and framework.
- 13 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Sensor Networks
- Vol. 22 (3) , 1-24
Open AccessMVSTA: Multi-View Spatio-Temporal Correlation Awareness Network for Traffic Data Imputation
Road traffic data imputation is an essential component of Intelligent Transportation Systems (ITS). However, the spatio-temporal characteristics in traffic data are complex and diverse, and existing methods are unable to comprehensively extract them. This article, through in-depth observations of traffic system dynamics, innovatively taxonomizes the complex spatio-temporal dependencies into four key dimensions: Geographical Spatial Correlations (GSC), Latent Spatial Correlations (LSC), Intra-sensor Temporal Correlations (ITC), and Cross-sensor Temporal Correlations (CTC). Motivated by this taxonomy, we aim at constructing a novel framework that holistically utilizes these spatio-temporal correlations to improve imputation accuracy. Subsequently, we propose a Multi-View Spatio-Temporal Correlation Awareness Network (MVSTA) for traffic data imputation, which incorporates two specifically designed modules: a Unified Spatial Correlation Awareness module (USCA) and a Collaborative Temporal Correlation Awareness module (CTCA). The USCA integrates GSC and LSC into a unified representation by jointly modeling physical proximity and data-driven dependencies. The CTCA collaboratively extracts ITC and CTC by capturing both local temporal patterns in individual sensors and interactive dynamics across different sensors. Extensive experiments on three real-world traffic datasets demonstrate that MVSTA significantly outperforms all baselines, validating the effectiveness of our proposed taxonomy and framework.
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 777-785
Unsupervised Subgraph Anomaly Detection Based on Pattern Collaboration
Subgraph Anomaly Detection (SAD) is crucial for identifying groups that deviate from the regular pattern within graphs, which benefits different domains such as financial fraud and network security. However, current studies rely on traditional node detection methods and fixed sampling strategies of subgraph structures, which makes it difficult to learn the pattern collaboration behavior of subgraphs. To address this limitation, this paper proposes a novel unsupervised framework named PC-SAD. The PC-SAD framework first employs an improved Graph AutoEncoder to identify core anomaly nodes by capturing multi-scale neighborhood information. Starting from these core anomaly nodes, we sample candidate subgraphs with path, tree, and cyclic structures, and enhance them according to the characteristics of the subgraph structures. Subsequently, candidate subgraphs are fed into the proposed Pattern Collaboration-based Graph Contrastive Learning method to generate collaborative pattern embeddings, thereby distinguishing anomaly subgraphs. The experimental results show that PC-SAD outperforms the state-of-the-art baseline methods on four benchmark datasets, which proves that PC-SAD is an effective solution to detect anomaly subgraphs.
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 777-785
Open AccessUnsupervised Subgraph Anomaly Detection Based on Pattern Collaboration
Subgraph Anomaly Detection (SAD) is crucial for identifying groups that deviate from the regular pattern within graphs, which benefits different domains such as financial fraud and network security. However, current studies rely on traditional node detection methods and fixed sampling strategies of subgraph structures, which makes it difficult to learn the pattern collaboration behavior of subgraphs. To address this limitation, this paper proposes a novel unsupervised framework named PC-SAD. The PC-SAD framework first employs an improved Graph AutoEncoder to identify core anomaly nodes by capturing multi-scale neighborhood information. Starting from these core anomaly nodes, we sample candidate subgraphs with path, tree, and cyclic structures, and enhance them according to the characteristics of the subgraph structures. Subsequently, candidate subgraphs are fed into the proposed Pattern Collaboration-based Graph Contrastive Learning method to generate collaborative pattern embeddings, thereby distinguishing anomaly subgraphs. The experimental results show that PC-SAD outperforms the state-of-the-art baseline methods on four benchmark datasets, which proves that PC-SAD is an effective solution to detect anomaly subgraphs.
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 5322-5333
Automated Model Selection for Multivariate Time Series Forecasting
Accurate multivariate time series forecasting (MTSF) is critical for intelligent web services in Web of Things. When confronted with unseen multivariate time series (MTS), the industry typically invests significant time and resources in training multiple models to identify the optimal model for deployment. This paper proposes a novel, efficient, and scalable MTSF model selection method that directly selects suitable MTSF methods based on data characteristics without extensive model training. Model selection is a core component of AutoML, which has made significant progress in recent years. However, existing methods incur high operational costs and cannot be directly applied to MTSF tasks. Moreover, there is a lack of a comprehensive and cohesive public time series library for MTSF model selection. To address these challenges, we compile the first large heterogeneous labeled MTSF model selection dataset, called the ModelPile, which covers 41 mainstream datasets across 11 domains. We then propose AutoMTSF, a large model-enabled model selection method that transforms the MTSF model selection problem into a time series classification problem and utilizes the ModelPile to unlock large-scale multi-dataset training. AutoMTSF first uses the pre-trained large model to encode raw MTS. Given the coarse-grained limitations of large model encoding, Recursive Temporal Pattern Feature (RTPF) is proposed to capture both fine-grained and global temporal feature evolution, thereby effectively mapping data characteristics to the MTSF method space. Experiments comparing AutoMTSF with 2 baselines, 17 MTSF methods, and 4 large time series models show that AutoMTSF outperforms state-of-the-art methods while maintaining comparable execution time. This work represents a critical step in validating the accuracy and efficiency of large model-enabled classification for MTSF.
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 5322-5333
Open AccessAutomated Model Selection for Multivariate Time Series Forecasting
Accurate multivariate time series forecasting (MTSF) is critical for intelligent web services in Web of Things. When confronted with unseen multivariate time series (MTS), the industry typically invests significant time and resources in training multiple models to identify the optimal model for deployment. This paper proposes a novel, efficient, and scalable MTSF model selection method that directly selects suitable MTSF methods based on data characteristics without extensive model training. Model selection is a core component of AutoML, which has made significant progress in recent years. However, existing methods incur high operational costs and cannot be directly applied to MTSF tasks. Moreover, there is a lack of a comprehensive and cohesive public time series library for MTSF model selection. To address these challenges, we compile the first large heterogeneous labeled MTSF model selection dataset, called the ModelPile, which covers 41 mainstream datasets across 11 domains. We then propose AutoMTSF, a large model-enabled model selection method that transforms the MTSF model selection problem into a time series classification problem and utilizes the ModelPile to unlock large-scale multi-dataset training. AutoMTSF first uses the pre-trained large model to encode raw MTS. Given the coarse-grained limitations of large model encoding, Recursive Temporal Pattern Feature (RTPF) is proposed to capture both fine-grained and global temporal feature evolution, thereby effectively mapping data characteristics to the MTSF method space. Experiments comparing AutoMTSF with 2 baselines, 17 MTSF methods, and 4 large time series models show that AutoMTSF outperforms state-of-the-art methods while maintaining comparable execution time. This work represents a critical step in validating the accuracy and efficiency of large model-enabled classification for MTSF.
- 1 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. PP (00983063) , 1
GCMT-Net: Generative Context-driven Multimodal Temporal Network for Video Anomaly Detection in Consumer Applications
Video constitutes a significant content modality within consumer electronics applications. Video Anomaly Detection (VAD) aims to automatically identify anomalous events deviating from normal patterns in consumer videos. This technology directly supports quality assurance and user experience optimization in consumer electronics products, while also enabling functionalities such as environmental safety monitoring for consumer electronics applications. However, traditional methods rely heavily on single-modal features, making it difficult to effectively utilize the complementary clues between different modalities. They have limitations and semantic gaps in explaining complex scenes and understanding the video context. Moreover, most methods overlook complex cross-modal correlations and fail to fully exploit the potential of multimodal semantic clues. To address these challenges, this work proposes an innovative Generative Context-driven Multimodal Temporal Network (GCMT-Net) to detect abnormal video for consumer electronic applications. The GCMT-Net introduces a context generation network that transforms visual context into rich textual descriptions, alleviating semantic gaps that are difficult to capture with traditional visual features. A novel multi-scale temporal module then integrates visual, auditory, and textual information, significantly enhancing the model’s ability to recognize complex anomalous events and improving its overall robustness. Furthermore, modal consistency learning ensures semantic alignment between modalities and consistent feature representations across scales. Experiments on four benchmark datasets demonstrate the superior performance of the proposed framework in addressing VAD challenges in complex consumer application scenarios.
- 1 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. PP (00983063) , 1
GCMT-Net: Generative Context-driven Multimodal Temporal Network for Video Anomaly Detection in Consumer Applications
Video constitutes a significant content modality within consumer electronics applications. Video Anomaly Detection (VAD) aims to automatically identify anomalous events deviating from normal patterns in consumer videos. This technology directly supports quality assurance and user experience optimization in consumer electronics products, while also enabling functionalities such as environmental safety monitoring for consumer electronics applications. However, traditional methods rely heavily on single-modal features, making it difficult to effectively utilize the complementary clues between different modalities. They have limitations and semantic gaps in explaining complex scenes and understanding the video context. Moreover, most methods overlook complex cross-modal correlations and fail to fully exploit the potential of multimodal semantic clues. To address these challenges, this work proposes an innovative Generative Context-driven Multimodal Temporal Network (GCMT-Net) to detect abnormal video for consumer electronic applications. The GCMT-Net introduces a context generation network that transforms visual context into rich textual descriptions, alleviating semantic gaps that are difficult to capture with traditional visual features. A novel multi-scale temporal module then integrates visual, auditory, and textual information, significantly enhancing the model’s ability to recognize complex anomalous events and improving its overall robustness. Furthermore, modal consistency learning ensures semantic alignment between modalities and consistent feature representations across scales. Experiments on four benchmark datasets demonstrate the superior performance of the proposed framework in addressing VAD challenges in complex consumer application scenarios.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (6) , 9300-9314
Long-Term Traffic Forecasting via Spatial-Temporal Wavelet Attention Network for Mobile IoT-Enabled Transportation Systems
Accurate long-term traffic forecasting improves traffic efficiency and safety in Intelligent Transportation Systems (ITS). However, previous methods typically overlook the complex mixed characteristics. So they struggle to capture the intricate features of newly added data effectively under distribution shifts, which exacerbates the complexity of spatial-temporal variations. Moreover, these methods fail to model global and local dynamic spatial correlations effectively, efficiently, and comprehensively. To mitigate these issues, this paper proposes an innovative Spatial-Temporal Wavelet Attention Network (STWAN). STWAN first decomposes traffic data into stable trends and fluctuating events, effectively dealing with the adverse effects of distribution shifts. Then, spatial-temporal encoder captures component-specific temporal variations and extracts dynamic global-local spatial correlations comprehensively with linear computational complexity. Additionally, transformer attention and forecasting decoder model the latent patterns, while trend-event fusion module integrates essential information for accurate forecasts.Comprehensive experiments across two real-world traffic forecasting tasks indicate that STWAN significantly outperforms state-of-the-art methods in terms of accuracy and robustness, attaining a maximum reduction of 5.54% in MAE and showcasing its robustness and broad applicability in long-term forecasting.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (6) , 9300-9314
Long-Term Traffic Forecasting via Spatial-Temporal Wavelet Attention Network for Mobile IoT-Enabled Transportation Systems
Accurate long-term traffic forecasting improves traffic efficiency and safety in Intelligent Transportation Systems (ITS). However, previous methods typically overlook the complex mixed characteristics. So they struggle to capture the intricate features of newly added data effectively under distribution shifts, which exacerbates the complexity of spatial-temporal variations. Moreover, these methods fail to model global and local dynamic spatial correlations effectively, efficiently, and comprehensively. To mitigate these issues, this paper proposes an innovative Spatial-Temporal Wavelet Attention Network (STWAN). STWAN first decomposes traffic data into stable trends and fluctuating events, effectively dealing with the adverse effects of distribution shifts. Then, spatial-temporal encoder captures component-specific temporal variations and extracts dynamic global-local spatial correlations comprehensively with linear computational complexity. Additionally, transformer attention and forecasting decoder model the latent patterns, while trend-event fusion module integrates essential information for accurate forecasts.Comprehensive experiments across two real-world traffic forecasting tasks indicate that STWAN significantly outperforms state-of-the-art methods in terms of accuracy and robustness, attaining a maximum reduction of 5.54% in MAE and showcasing its robustness and broad applicability in long-term forecasting.
- 4 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 13, 4562-4577
DRL-Based Weakly Supervised Traffic Anomaly Detection for IoT Networks
With the proliferation of Internet of Things (IoT) devices, IoT systems face increasingly severe threats from large-scale cyberattacks during network communications. To effectively recognize potential network threats and ensure the security of IoT, network traffic anomaly detection has been widely studied. Due to the high cost of manual labeling in real-world scenarios, only limited network traffic data is explicitly labeled as abnormal or normal. However, most existing anomaly detection methods struggle to enhance detection accuracy with minimal supervision and perform ineffectively at identifying potential unknown anomalies in unlabeled data. To address these limitations, this paper proposes WADE, a Weakly supervised Anomaly DEtection method based on Deep Reinforcement Learning (DRL). WADE enables simultaneous identification of known and unknown anomaly traffic with limited labeled data. Specifically, it incorporates a Dueling Q-network architecture and introduces a novel reward optimization mechanism that: 1) strengthens feature extraction from unlabeled data and 2) elevates decision accuracy via dynamic reward adaptation. Extensive experiments on IoT traffic datasets demonstrate that WADE outperforms four peer weakly supervised methods, achieving performance improvements of 12.2$\%$ in Area Under the Precision-Recall Curve (AUC-PR) and 6.6$\%$ in Area Under the Receiver Operating Characteristic Curve (AUC-ROC), which validate the effectiveness of WADE in safeguarding IoT systems.
- 4 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 13, 4562-4577
DRL-Based Weakly Supervised Traffic Anomaly Detection for IoT Networks
With the proliferation of Internet of Things (IoT) devices, IoT systems face increasingly severe threats from large-scale cyberattacks during network communications. To effectively recognize potential network threats and ensure the security of IoT, network traffic anomaly detection has been widely studied. Due to the high cost of manual labeling in real-world scenarios, only limited network traffic data is explicitly labeled as abnormal or normal. However, most existing anomaly detection methods struggle to enhance detection accuracy with minimal supervision and perform ineffectively at identifying potential unknown anomalies in unlabeled data. To address these limitations, this paper proposes WADE, a Weakly supervised Anomaly DEtection method based on Deep Reinforcement Learning (DRL). WADE enables simultaneous identification of known and unknown anomaly traffic with limited labeled data. Specifically, it incorporates a Dueling Q-network architecture and introduces a novel reward optimization mechanism that: 1) strengthens feature extraction from unlabeled data and 2) elevates decision accuracy via dynamic reward adaptation. Extensive experiments on IoT traffic datasets demonstrate that WADE outperforms four peer weakly supervised methods, achieving performance improvements of 12.2$\%$ in Area Under the Precision-Recall Curve (AUC-PR) and 6.6$\%$ in Area Under the Receiver Operating Characteristic Curve (AUC-ROC), which validate the effectiveness of WADE in safeguarding IoT systems.