Nan Zhang
Research Subjects
Nan Zhang has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Xiaolong Yang
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing 100083, China
4 shared publications - Lisha Shuai
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing 100083, China
3 shared publications - Gaoya Dong
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing 100083, China
2 shared publications - Haokun Zhang
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, China
2 shared publications - Shaofeng Tan
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, China
2 shared publications - Cai Liu
State Grid Information & Telecommunication Group Co. Ltd, Beijing 102211, China
1 shared publication
Latest Publications
- Open Access
- 27 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 4777-4794
PEEL: A Poisoning-Exposing Encoding Mechanism for Local Differential Privacy
Local Differential Privacy (LDP) has emerged as a dominant privacy paradigm in the Internet of Things (IoT) due to its lightweight, decentralized and scalable nature. However, the indistinguishability guarantee inherent to LDP mechanisms inadvertently masks adversarial manipulation, rendering poisoned data difficult to differentiate from legitimate perturbations. Existing poisoning defense mechanisms either incur prohibitive overheads incompatible with resource-constrained IoT devices or rely on rigid domain priors. To address this, we propose PEEL, a Poisoning-Exposing Encoding mechanism for LDP, which reformulates poisoning exposure as a geometric structural verification problem within the encoding subspace. Departing from probabilistic statistical filtering, PEEL exploits the rank disparity between legitimate sparse perturbations and arbitrary adversarial injections. By projecting LDP outputs onto a low-rank structural support, the mechanism forces poisoning attempts to manifest as deterministic reconstruction residuals, thereby exposing structure-violating output and rule poisoning attacks. Theoretical analysis confirms that PEEL, as a post-randomization encoding layer, preserves the unbiasedness and statistical accuracy of the underlying LDP protocols, while deterministically exposing structure-violating output and rule poisoning attacks. Extensive evaluations across six protocols and two real-world datasets confirm that PEEL reduces client-side overheads by orders of magnitude compared to cryptographic baselines, while maintaining F1 scores exceeding 99.8% against structure-violating output and rule poisoning attacks.
- 27 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 4777-4794
Open AccessPEEL: A Poisoning-Exposing Encoding Mechanism for Local Differential Privacy
Local Differential Privacy (LDP) has emerged as a dominant privacy paradigm in the Internet of Things (IoT) due to its lightweight, decentralized and scalable nature. However, the indistinguishability guarantee inherent to LDP mechanisms inadvertently masks adversarial manipulation, rendering poisoned data difficult to differentiate from legitimate perturbations. Existing poisoning defense mechanisms either incur prohibitive overheads incompatible with resource-constrained IoT devices or rely on rigid domain priors. To address this, we propose PEEL, a Poisoning-Exposing Encoding mechanism for LDP, which reformulates poisoning exposure as a geometric structural verification problem within the encoding subspace. Departing from probabilistic statistical filtering, PEEL exploits the rank disparity between legitimate sparse perturbations and arbitrary adversarial injections. By projecting LDP outputs onto a low-rank structural support, the mechanism forces poisoning attempts to manifest as deterministic reconstruction residuals, thereby exposing structure-violating output and rule poisoning attacks. Theoretical analysis confirms that PEEL, as a post-randomization encoding layer, preserves the unbiasedness and statistical accuracy of the underlying LDP protocols, while deterministically exposing structure-violating output and rule poisoning attacks. Extensive evaluations across six protocols and two real-world datasets confirm that PEEL reduces client-side overheads by orders of magnitude compared to cryptographic baselines, while maintaining F1 scores exceeding 99.8% against structure-violating output and rule poisoning attacks.
- 17 January 2026
- Published by Elsevier in Computer Standards & Interfaces
- Vol. 97
Privacy-preserving kNN classification for cross-platform electric vehicle data analytics
- 17 January 2026
- Published by Elsevier in Computer Standards & Interfaces
- Vol. 97
Privacy-preserving kNN classification for cross-platform electric vehicle data analytics
- Open Access
PEEL: A Poisoning-Exposing Encoding Theoretical Framework for Local Differential Privacy
Local Differential Privacy (LDP) is a widely adopted privacy-protection model in the Internet of Things (IoT) due to its lightweight, decentralized, and scalable nature. However, it is vulnerable to poisoning attacks, and existing defenses either incur prohibitive resource overheads or rely on domain-specific prior knowledge, limiting their practical deployment. To address these limitations, we propose PEEL, a Poisoning-Exposing Encoding theoretical framework for LDP, which departs from resource- or prior-dependent countermeasures and instead leverages the inherent structural consistency of LDP-perturbed data. As a non-intrusive post-processing module, PEEL amplifies stealthy poisoning effects by re-encoding LDP-perturbed data via sparsification, normalization, and low-rank projection, thereby revealing both output and rule poisoning attacks through structural inconsistencies in the reconstructed space. Theoretical analysis proves that PEEL, integrated with LDP, retains unbiasedness and statistical accuracy, while being robust to expose both output and rule poisoning attacks. Moreover, evaluation results show that LDP-integrated PEEL not only outperforms four state-of-the-art defenses in terms of poisoning exposure accuracy but also significantly reduces client-side computational costs, making it highly suitable for large-scale IoT deployments.
Open AccessPEEL: A Poisoning-Exposing Encoding Theoretical Framework for Local Differential Privacy
Local Differential Privacy (LDP) is a widely adopted privacy-protection model in the Internet of Things (IoT) due to its lightweight, decentralized, and scalable nature. However, it is vulnerable to poisoning attacks, and existing defenses either incur prohibitive resource overheads or rely on domain-specific prior knowledge, limiting their practical deployment. To address these limitations, we propose PEEL, a Poisoning-Exposing Encoding theoretical framework for LDP, which departs from resource- or prior-dependent countermeasures and instead leverages the inherent structural consistency of LDP-perturbed data. As a non-intrusive post-processing module, PEEL amplifies stealthy poisoning effects by re-encoding LDP-perturbed data via sparsification, normalization, and low-rank projection, thereby revealing both output and rule poisoning attacks through structural inconsistencies in the reconstructed space. Theoretical analysis proves that PEEL, integrated with LDP, retains unbiasedness and statistical accuracy, while being robust to expose both output and rule poisoning attacks. Moreover, evaluation results show that LDP-integrated PEEL not only outperforms four state-of-the-art defenses in terms of poisoning exposure accuracy but also significantly reduces client-side computational costs, making it highly suitable for large-scale IoT deployments.
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (19) , 39991-40007
PoisonCatcher: Revealing and Identifying LDP Poisoning Attacks in IIoT
Local differential privacy (LDP), a robust privacy-protection model, is widely adopted in the Industrial Internet of Things (IIoT) due to its lightweight, decentralized, and scalable. However, its perturbation-based privacy-protection mechanism hinders distinguishing between any two data, thereby facilitating LDP poisoning attacks. The exposed physical-layer vulnerabilities and resource-constrained prevalent at the IIoT edge not only facilitate such attacks but also render existing LDP poisoning defenses, all of which are deployed at the edge and rely on ample resources, impractical. This work proposes a LDP poisoning defense for IIoT in the resource-rich aggregator. We first reveal key poisoning attack modes occurring within the LDP-utilized IIoT data-collection process, detailing how IIoT vulnerabilities enable attacks, and then formulate a general attack model and derive the poisoned data’s indistinguishability. This work subsequently analyzes the poisoning impacts on aggregated data based on industrial process correlation, revealing the distortion of statistical query result’ temporal similarity and the resulting disruption of interattribute correlation, and uncovering the intriguing paradox that adversaries’ attempts to stabilize their poisoning actions for stealth are difficult to maintain. Given these findings, we propose PoisonCatcher, a solution for identifying poisoned data, which includes time-series detectors based on temporal similarity, attribute correlation, and pattern stability metrics to detect poisoned attributes, and a latent-bias feature miner for identifying poisons. Experiments on two representative real-world datasets indicate that PoisonCatcher successfully identifies poisoned data, demonstrating robust identification capabilities with F2 scores above 90.7% under various attack settings.
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (19) , 39991-40007
PoisonCatcher: Revealing and Identifying LDP Poisoning Attacks in IIoT
Local differential privacy (LDP), a robust privacy-protection model, is widely adopted in the Industrial Internet of Things (IIoT) due to its lightweight, decentralized, and scalable. However, its perturbation-based privacy-protection mechanism hinders distinguishing between any two data, thereby facilitating LDP poisoning attacks. The exposed physical-layer vulnerabilities and resource-constrained prevalent at the IIoT edge not only facilitate such attacks but also render existing LDP poisoning defenses, all of which are deployed at the edge and rely on ample resources, impractical. This work proposes a LDP poisoning defense for IIoT in the resource-rich aggregator. We first reveal key poisoning attack modes occurring within the LDP-utilized IIoT data-collection process, detailing how IIoT vulnerabilities enable attacks, and then formulate a general attack model and derive the poisoned data’s indistinguishability. This work subsequently analyzes the poisoning impacts on aggregated data based on industrial process correlation, revealing the distortion of statistical query result’ temporal similarity and the resulting disruption of interattribute correlation, and uncovering the intriguing paradox that adversaries’ attempts to stabilize their poisoning actions for stealth are difficult to maintain. Given these findings, we propose PoisonCatcher, a solution for identifying poisoned data, which includes time-series detectors based on temporal similarity, attribute correlation, and pattern stability metrics to detect poisoned attributes, and a latent-bias feature miner for identifying poisons. Experiments on two representative real-world datasets indicate that PoisonCatcher successfully identifies poisoned data, demonstrating robust identification capabilities with F2 scores above 90.7% under various attack settings.
- Open Access
- 24 January 2022
- Published by Hindawi Limited in Wireless Communications and Mobile Computing
- Vol. 2022, 1-11
Deep Learning with Multisource Data Fusion in Electricity Internet of Things for Electricity Price Forecast
More and more IoT (Internet of Thing) devices have been connected to our lives in recent years, making life more convenient. Many countries are also making use of Internet of Thing technology to carry out intelligent electricity network reform. One of the reform goals is balancing the supply and demand of electricity, which has become a top priority. Balancing electricity supply and demand through real-time electricity prices has become an effective way. However, using traditional machine learning models for real-time electricity price prediction requires complex feature engineering, and the results are not satisfactory. Also, the mainstream fusion methods use data-level fusion, which will put very high pressure on communication bandwidth and computer resources. In this paper, an LSTM- (long short-term memory-) based decision level fusion of multisource data is proposed and applied for real-time electricity price prediction on actual electricity price datasets. The method solves the difficulties of traditional machine learning models in dealing with complex nonlinear problems. It achieves local asynchronous processing of multisource data through decision-level fusion, reducing the requirement for bandwidth resources and providing perfect results in real-time electricity price prediction. The experimental results show that the prediction accuracy of the decision fusion prediction model based on LSTM is higher than that of the linear regression algorithm.
- 24 January 2022
- Published by Hindawi Limited in Wireless Communications and Mobile Computing
- Vol. 2022, 1-11
Open AccessDeep Learning with Multisource Data Fusion in Electricity Internet of Things for Electricity Price Forecast
More and more IoT (Internet of Thing) devices have been connected to our lives in recent years, making life more convenient. Many countries are also making use of Internet of Thing technology to carry out intelligent electricity network reform. One of the reform goals is balancing the supply and demand of electricity, which has become a top priority. Balancing electricity supply and demand through real-time electricity prices has become an effective way. However, using traditional machine learning models for real-time electricity price prediction requires complex feature engineering, and the results are not satisfactory. Also, the mainstream fusion methods use data-level fusion, which will put very high pressure on communication bandwidth and computer resources. In this paper, an LSTM- (long short-term memory-) based decision level fusion of multisource data is proposed and applied for real-time electricity price prediction on actual electricity price datasets. The method solves the difficulties of traditional machine learning models in dealing with complex nonlinear problems. It achieves local asynchronous processing of multisource data through decision-level fusion, reducing the requirement for bandwidth resources and providing perfect results in real-time electricity price prediction. The experimental results show that the prediction accuracy of the decision fusion prediction model based on LSTM is higher than that of the linear regression algorithm.
- Open Access
- 7 September 2017
- Published by Wiley in International Journal of Communication Systems
- Vol. 31 (1)
A genetic algorithm‐based task scheduling for cloud resource crowd‐funding model
Summary: With the rapid development of cloud computing, the number of cloud users is growing exponentially. Data centers have come under great pressure, and the problem of power consumption has become increasingly prominent. However, many idle resources that are geographically distributed in the network can be used as resource providers for cloud tasks. These distributed resources may not be able to support the resource‐intensive applications alone because of their limited capacity; however, the capacity will be considerably increased if they can cooperate with each other and share resources. Therefore, in this paper, a new resource‐providing model called “crowd‐funding” is proposed. In the crowd‐funding model, idle resources can be collected to form a virtual resource pool for providing cloud services. Based on this model, a new task scheduling algorithm is proposed, RC‐GA (genetic algorithm for task scheduling based on a resource crowd‐funding model). For crowd‐funding, the resources come from different heterogeneous devices, so the resource stability should be considered different. The scheduling targets of the RC‐GA are designed to increase the stability of task execution and reduce power consumption at the same time. In addition, to reduce random errors in the evolution process, the roulette wheel selection operator of the genetic algorithm is improved. The experiment shows that the RC‐GA can achieve good results.
- 7 September 2017
- Published by Wiley in International Journal of Communication Systems
- Vol. 31 (1)
Open AccessA genetic algorithm‐based task scheduling for cloud resource crowd‐funding model
Summary: With the rapid development of cloud computing, the number of cloud users is growing exponentially. Data centers have come under great pressure, and the problem of power consumption has become increasingly prominent. However, many idle resources that are geographically distributed in the network can be used as resource providers for cloud tasks. These distributed resources may not be able to support the resource‐intensive applications alone because of their limited capacity; however, the capacity will be considerably increased if they can cooperate with each other and share resources. Therefore, in this paper, a new resource‐providing model called “crowd‐funding” is proposed. In the crowd‐funding model, idle resources can be collected to form a virtual resource pool for providing cloud services. Based on this model, a new task scheduling algorithm is proposed, RC‐GA (genetic algorithm for task scheduling based on a resource crowd‐funding model). For crowd‐funding, the resources come from different heterogeneous devices, so the resource stability should be considered different. The scheduling targets of the RC‐GA are designed to increase the stability of task execution and reduce power consumption at the same time. In addition, to reduce random errors in the evolution process, the roulette wheel selection operator of the genetic algorithm is improved. The experiment shows that the RC‐GA can achieve good results.