Pub Date : 2024-04-18DOI: 10.1109/TSUSC.2024.3387727
Kaikai Qiao;Ben Ma;Lidan Wang;Shukai Duan
In recent years, memristors have attracted much attention in the fields of nonvolatile memory, logic operation and neuromorphic computing. As a new type of two-terminal passive electronic component similar to sandwich structure, its main resistance mechanism is the formation and fracture of metal or oxygen vacancy conductive filaments. Traditional clustering algorithms own strong sensitivity to different parameter selection, including partition clustering algorithm and density clustering algorithm. In view of the non-volatile characteristics of memristor and the In-memory computing characteristics of memristive circuit, this paper designs a new memristive clustering paradigm, and further verifies the feasibility and effectiveness of the proposed analog circuit to improve the performance of clustering parameters by exploring the data mining and image segmentation problems of these two types of clustering algorithms.
{"title":"Memristive Clustering: A Novel Sustainable Parameter Selection Based on Memristive Circuit Model","authors":"Kaikai Qiao;Ben Ma;Lidan Wang;Shukai Duan","doi":"10.1109/TSUSC.2024.3387727","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3387727","url":null,"abstract":"In recent years, memristors have attracted much attention in the fields of nonvolatile memory, logic operation and neuromorphic computing. As a new type of two-terminal passive electronic component similar to sandwich structure, its main resistance mechanism is the formation and fracture of metal or oxygen vacancy conductive filaments. Traditional clustering algorithms own strong sensitivity to different parameter selection, including partition clustering algorithm and density clustering algorithm. In view of the non-volatile characteristics of memristor and the In-memory computing characteristics of memristive circuit, this paper designs a new memristive clustering paradigm, and further verifies the feasibility and effectiveness of the proposed analog circuit to improve the performance of clustering parameters by exploring the data mining and image segmentation problems of these two types of clustering algorithms.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"10 1","pages":"18-27"},"PeriodicalIF":3.0,"publicationDate":"2024-04-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143184025","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Efficient and accurate identification of network anomalies is significant to network security systems. It is highly challenging to detect abnormal behaviors in the increasing network data accurately. Currently, classification methods based on feature extraction of autoencoders have been proven to be suitable for network anomaly detection. However, traditional detection models with autoencoders have unsatisfying detection accuracy in the face of massive network features. In addition, the hyperparameter optimization of their models cannot be effectively solved. In this letter, based on the improvement of variational autoencoders, stacked sparse shrink variational autoencoders (S3VAEs) are designed. In addition, an Unbalanced XGBoost classifier based on Genetic simulated annealing particle swarm optimization (UXG) is proposed. Finally, the feature extractor of S3VAEs is combined with the UXG classifier, and the anomaly detection model is obtained. Experimental results based on four real-life data sets demonstrate that the proposed anomaly detection model achieves higher classification accuracy and F1 than several state-of-the-art algorithms.
{"title":"Network Anomaly Detection With Stacked Sparse Shrink Variational Autoencoders and Unbalanced XGBoost","authors":"Jing Bi;Ziyue Guan;Haitao Yuan;Jinhong Yang;Jia Zhang","doi":"10.1109/TSUSC.2024.3390003","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3390003","url":null,"abstract":"Efficient and accurate identification of network anomalies is significant to network security systems. It is highly challenging to detect abnormal behaviors in the increasing network data accurately. Currently, classification methods based on feature extraction of autoencoders have been proven to be suitable for network anomaly detection. However, traditional detection models with autoencoders have unsatisfying detection accuracy in the face of massive network features. In addition, the hyperparameter optimization of their models cannot be effectively solved. In this letter, based on the improvement of variational autoencoders, stacked sparse shrink variational autoencoders (S3VAEs) are designed. In addition, an <underline>U</u>nbalanced <underline>X</u>GBoost classifier based on <underline>G</u>enetic simulated annealing particle swarm optimization (UXG) is proposed. Finally, the feature extractor of S3VAEs is combined with the UXG classifier, and the anomaly detection model is obtained. Experimental results based on four real-life data sets demonstrate that the proposed anomaly detection model achieves higher classification accuracy and F1 than several state-of-the-art algorithms.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"10 1","pages":"28-38"},"PeriodicalIF":3.0,"publicationDate":"2024-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143184026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Federated learning (FL) is a distributed machine learning paradigm that addresses the challenges of privacy leakage and data silos by collaboratively training the global model through parameter exchange, rather than data, between the central server and local clients. However, recent researches highlight the vulnerability of FL to gradient leakage attacks where adversaries exploit shared parameters from clients to reconstruct sensitive training data. Differential privacy (DP) effectively mitigates this threat by adding noise to shared parameters, yet introduces a trade-off between privacy and accuracy in FL. To better balance the privacy and accuracy, in this paper we propose a staged noise perturbation strategy, called alternating noise permutation (ANP), from a novel perspective. ANP adds Gaussian-distributed random noise to model parameters during the critical learning period of FL, following DP principles. While in non-critical learning period, ANP alternately permutes the noise during odd and even communication rounds, achieving near mutual cancellation and mitigating the negative impact. Experimental results across three datasets and two neural networks under both independent identical distribution (IID) and NonIID scenarios demonstrate that ANP significantly improves classification accuracy and exhibits robustness against gradient leakage attack, ensuring the effectiveness of FL for secure and accurate collaborative model training.
{"title":"Staged Noise Perturbation for Privacy-Preserving Federated Learning","authors":"Zhe Li;Honglong Chen;Yudong Gao;Zhichen Ni;Huansheng Xue;Huajie Shao","doi":"10.1109/TSUSC.2024.3381812","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3381812","url":null,"abstract":"Federated learning (FL) is a distributed machine learning paradigm that addresses the challenges of privacy leakage and data silos by collaboratively training the global model through parameter exchange, rather than data, between the central server and local clients. However, recent researches highlight the vulnerability of FL to gradient leakage attacks where adversaries exploit shared parameters from clients to reconstruct sensitive training data. Differential privacy (DP) effectively mitigates this threat by adding noise to shared parameters, yet introduces a trade-off between privacy and accuracy in FL. To better balance the privacy and accuracy, in this paper we propose a staged noise perturbation strategy, called alternating noise permutation (ANP), from a novel perspective. ANP adds Gaussian-distributed random noise to model parameters during the critical learning period of FL, following DP principles. While in non-critical learning period, ANP alternately permutes the noise during odd and even communication rounds, achieving near mutual cancellation and mitigating the negative impact. Experimental results across three datasets and two neural networks under both independent identical distribution (IID) and NonIID scenarios demonstrate that ANP significantly improves classification accuracy and exhibits robustness against gradient leakage attack, ensuring the effectiveness of FL for secure and accurate collaborative model training.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"9 6","pages":"936-947"},"PeriodicalIF":3.0,"publicationDate":"2024-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142810512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-27DOI: 10.1109/TSUSC.2024.3382157
Xian Zhang;Guoqing Xiao;Mingxing Duan;Yuedan Chen;Kenli Li
Convolutional neural networks (CNNs) are widely utilized in intelligent edge computing applications such as computational vision and image processing. However, as the number of layers of the CNN model increases, the number of parameters and computations gets larger, making it increasingly challenging to accelerate in edge computing applications. To effectively adapt to the tradeoff between the speed and accuracy of CNNs inference for smart applications. This paper proposes an FPGA-based adaptive CNNs inference accelerator synergistically utilizing filter pruning, fixed-point parameter quantization, and multi-computing unit parallelism called APPQ-CNN. First, the article devises a hybrid pruning algorithm based on the L1-norm and APoZ to measure the filter impact degree and a configurable parameter quantization fixed-point computing architecture instead of floating-point architecture. Then, design a cascade of the CNN pipelined kernel architecture and configurable multiple computation units. Finally, conduct extensive performance exploration and comparison experiments on various real and synthetic datasets. With negligible accuracy loss, the speed performance of our accelerator APPQ-CNN compares with current state-of-the-art FPGA-based accelerators PipeCNN and OctCNN by 2.15× and 1.91×, respectively. Furthermore, APPQ-CNN provides settable fixed-point quantization bit-width parameters, filter pruning rate, and multiple computation unit counts to cope with practical application performance requirements in edge computing.
{"title":"APPQ-CNN: An Adaptive CNNs Inference Accelerator for Synergistically Exploiting Pruning and Quantization Based on FPGA","authors":"Xian Zhang;Guoqing Xiao;Mingxing Duan;Yuedan Chen;Kenli Li","doi":"10.1109/TSUSC.2024.3382157","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3382157","url":null,"abstract":"Convolutional neural networks (CNNs) are widely utilized in intelligent edge computing applications such as computational vision and image processing. However, as the number of layers of the CNN model increases, the number of parameters and computations gets larger, making it increasingly challenging to accelerate in edge computing applications. To effectively adapt to the tradeoff between the speed and accuracy of CNNs inference for smart applications. This paper proposes an FPGA-based adaptive CNNs inference accelerator synergistically utilizing filter pruning, fixed-point parameter quantization, and multi-computing unit parallelism called APPQ-CNN. First, the article devises a hybrid pruning algorithm based on the L1-norm and APoZ to measure the filter impact degree and a configurable parameter quantization fixed-point computing architecture instead of floating-point architecture. Then, design a cascade of the CNN pipelined kernel architecture and configurable multiple computation units. Finally, conduct extensive performance exploration and comparison experiments on various real and synthetic datasets. With negligible accuracy loss, the speed performance of our accelerator APPQ-CNN compares with current state-of-the-art FPGA-based accelerators PipeCNN and OctCNN by 2.15× and 1.91×, respectively. Furthermore, APPQ-CNN provides settable fixed-point quantization bit-width parameters, filter pruning rate, and multiple computation unit counts to cope with practical application performance requirements in edge computing.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"9 6","pages":"874-888"},"PeriodicalIF":3.0,"publicationDate":"2024-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142810536","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The rapid advancement of mobile edge computing (MEC) has revolutionized the distributed computing landscape. With the help of MEC, the traditional centralized cloud computing architecture can be extended to the edge of networks, enabling real-time processing of resources and time-sensitive applications. Nevertheless, the problem of efficiently assigning the services to the computing resources is a challenging and prevalent issue due to the dynamic and distributed nature of the edge network's architecture. Thus, we require intelligent real-time decision-making and effective optimization algorithms to allocate resources, such as network bandwidth, memory, and CPU. This paper proposes an MEC architecture to allocate the resources in the network to optimize the quality of services (QoS). In this regard, the resource allocation problem is formulated as a bi-objective optimization problem, including minimizing cost and energy with quality and deadline constraints. A hybrid cascading-based meta-heuristic called GA-PSO is embedded with the proposed MEC architecture to achieve these objectives. Finally, it is compared with three existing approaches to establish its efficacy. The experimental results report statistically better cost and energy in all the considered instances, making it practical and validating its effectiveness.
{"title":"Deadline-Aware Cost and Energy Efficient Offloading in Mobile Edge Computing","authors":"Mohit Kumar;Avadh Kishor;Pramod Kumar Singh;Kalka Dubey","doi":"10.1109/TSUSC.2024.3381841","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3381841","url":null,"abstract":"The rapid advancement of mobile edge computing (MEC) has revolutionized the distributed computing landscape. With the help of MEC, the traditional centralized cloud computing architecture can be extended to the edge of networks, enabling real-time processing of resources and time-sensitive applications. Nevertheless, the problem of efficiently assigning the services to the computing resources is a challenging and prevalent issue due to the dynamic and distributed nature of the edge network's architecture. Thus, we require intelligent real-time decision-making and effective optimization algorithms to allocate resources, such as network bandwidth, memory, and CPU. This paper proposes an MEC architecture to allocate the resources in the network to optimize the quality of services (QoS). In this regard, the resource allocation problem is formulated as a bi-objective optimization problem, including minimizing cost and energy with quality and deadline constraints. A hybrid cascading-based meta-heuristic called GA-PSO is embedded with the proposed MEC architecture to achieve these objectives. Finally, it is compared with three existing approaches to establish its efficacy. The experimental results report statistically better cost and energy in all the considered instances, making it practical and validating its effectiveness.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"9 5","pages":"778-789"},"PeriodicalIF":3.0,"publicationDate":"2024-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142397224","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Wireless Power Transfer (WPT) is a disruptive technology that allows wireless energy provisioning for energy-limited IoT devices, thus decreasing the over-reliance on batteries and wires. WPT could replace conventional energy provisioning (e.g., energy harvesting) and expand to be deployed in many of our daily-life applications, including but not limited to healthcare, transportation, automation, and smart cities. As a new rising technology, WPT has attracted many researchers from academia and industry about WPT technologies and wireless charging scheduling algorithms. Therefore, in this paper, we review the most recent studies related to WPT, including classifications, advantages, disadvantages, and main domains of application. Furthermore, we review the recently designed wireless charging scheduling algorithms (schemes) for wireless sensor networks. Our study provides a detailed survey of wireless charging scheduling schemes covering the main scheme classifications, evaluation metrics, application domains, advantages, and disadvantages of each charging scheme. We further summarize trends and opportunities for applying WPT at some intersections.
{"title":"Wireless Power Transfer Technologies, Applications, and Future Trends: A Review","authors":"Aisha Alabsi;Ammar Hawbani;Xingfu Wang;Ahmed Al-Dubai;Jiankun Hu;Samah Abdel Aziz;Santosh Kumar;Liang Zhao;Alexey V. Shvetsov;Saeed Hamood Alsamhi","doi":"10.1109/TSUSC.2024.3380607","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3380607","url":null,"abstract":"Wireless Power Transfer (WPT) is a disruptive technology that allows wireless energy provisioning for energy-limited IoT devices, thus decreasing the over-reliance on batteries and wires. WPT could replace conventional energy provisioning (e.g., energy harvesting) and expand to be deployed in many of our daily-life applications, including but not limited to healthcare, transportation, automation, and smart cities. As a new rising technology, WPT has attracted many researchers from academia and industry about WPT technologies and wireless charging scheduling algorithms. Therefore, in this paper, we review the most recent studies related to WPT, including classifications, advantages, disadvantages, and main domains of application. Furthermore, we review the recently designed wireless charging scheduling algorithms (schemes) for wireless sensor networks. Our study provides a detailed survey of wireless charging scheduling schemes covering the main scheme classifications, evaluation metrics, application domains, advantages, and disadvantages of each charging scheme. We further summarize trends and opportunities for applying WPT at some intersections.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"10 1","pages":"1-17"},"PeriodicalIF":3.0,"publicationDate":"2024-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143184024","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-22DOI: 10.1109/TSUSC.2024.3395350
Na Hu;Wei Liang;Dafang Zhang;Kun Xie;Kuanching Li;Albert Y. Zomaya
Traffic prediction is crucial for intelligent transportation systems, assisting in making travel decisions, minimizing traffic congestion, and improving traffic operation efficiency. Although effective, existing centralized traffic prediction methods have privacy leakage risks. Federated learning-based traffic prediction methods keep raw data local and train the global model in a distributed way, thus preserving data privacy. Nevertheless, the spatial correlations between local clients will be broken as data exchange between local clients is not allowed in federated learning, leading to missing spatial information and inferior prediction accuracy. To this end, we propose a federated graph neural network with spatial information completion (FedGCN) for privacy-preserving traffic prediction by adopting a federated learning scheme to protect confidentiality and presenting a mending graph convolutional neural network to mend the missing spatial information during capturing spatial dependency to improve prediction accuracy. To complete the missing spatial information efficiently and capture the client-specific spatial pattern, we design a personalized training scheme for the mending graph neural network, reducing communication overhead. The experiments on four public traffic datasets demonstrate that the proposed model outperforms the best baseline with a ratio of 3.82%, 1.82%, 2.13%, and 1.49% in terms of absolute mean error while preserving privacy.
{"title":"FedGCN: A Federated Graph Convolutional Network for Privacy-Preserving Traffic Prediction","authors":"Na Hu;Wei Liang;Dafang Zhang;Kun Xie;Kuanching Li;Albert Y. Zomaya","doi":"10.1109/TSUSC.2024.3395350","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3395350","url":null,"abstract":"Traffic prediction is crucial for intelligent transportation systems, assisting in making travel decisions, minimizing traffic congestion, and improving traffic operation efficiency. Although effective, existing centralized traffic prediction methods have privacy leakage risks. Federated learning-based traffic prediction methods keep raw data local and train the global model in a distributed way, thus preserving data privacy. Nevertheless, the spatial correlations between local clients will be broken as data exchange between local clients is not allowed in federated learning, leading to missing spatial information and inferior prediction accuracy. To this end, we propose a federated graph neural network with spatial information completion (FedGCN) for privacy-preserving traffic prediction by adopting a federated learning scheme to protect confidentiality and presenting a mending graph convolutional neural network to mend the missing spatial information during capturing spatial dependency to improve prediction accuracy. To complete the missing spatial information efficiently and capture the client-specific spatial pattern, we design a personalized training scheme for the mending graph neural network, reducing communication overhead. The experiments on four public traffic datasets demonstrate that the proposed model outperforms the best baseline with a ratio of 3.82%, 1.82%, 2.13%, and 1.49% in terms of absolute mean error while preserving privacy.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"9 6","pages":"925-935"},"PeriodicalIF":3.0,"publicationDate":"2024-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142810511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Federated Learning (FL), as a distributed machine learning technique, has promise for training models with distributed data in Artificial Intelligence of Things (AIoT). However, FL is vulnerable to Byzantine attacks from diverse participants. While numerous Byzantine-robust FL solutions have been proposed, most of them involve deploying defenses at either the aggregation server or the participant level, significantly impacting the original FL process. Moreover, it will bring extra computational burden to the server or the participant, which is especially unsuitable for the resource-constrained AIoT domain. To resolve the aforementioned concerns, we propose FL-Auditor, a Byzantine-robust FL approach based on public auditing. Its core idea is to use a Third-Party Auditor (TPA) to audit samples from the FL training process, analyzing the trustworthiness of different participants, thereby helping FL obtain a more robust global model. In addition, we also design a lazy update mechanism to reduce the negative impact of sampling audit on the performance of the global model. Extensive experiments have demonstrated the effectiveness of our FL-Auditor in terms of accuracy, robustness against attacks, and flexibility. In particular, compared to the existing method, our FL-Auditor significantly reduces the computation time on the aggregation server by 8×-17×.
{"title":"Using Third-Party Auditor to Help Federated Learning: An Efficient Byzantine-Robust Federated Learning","authors":"Zhuangzhuang Zhang;Libing Wu;Debiao He;Jianxin Li;Na Lu;Xuejiang Wei","doi":"10.1109/TSUSC.2024.3379440","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3379440","url":null,"abstract":"Federated Learning (FL), as a distributed machine learning technique, has promise for training models with distributed data in Artificial Intelligence of Things (AIoT). However, FL is vulnerable to Byzantine attacks from diverse participants. While numerous Byzantine-robust FL solutions have been proposed, most of them involve deploying defenses at either the aggregation server or the participant level, significantly impacting the original FL process. Moreover, it will bring extra computational burden to the server or the participant, which is especially unsuitable for the resource-constrained AIoT domain. To resolve the aforementioned concerns, we propose FL-Auditor, a Byzantine-robust FL approach based on public auditing. Its core idea is to use a Third-Party Auditor (TPA) to audit samples from the FL training process, analyzing the trustworthiness of different participants, thereby helping FL obtain a more robust global model. In addition, we also design a lazy update mechanism to reduce the negative impact of sampling audit on the performance of the global model. Extensive experiments have demonstrated the effectiveness of our FL-Auditor in terms of accuracy, robustness against attacks, and flexibility. In particular, compared to the existing method, our FL-Auditor significantly reduces the computation time on the aggregation server by 8×-17×.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"9 6","pages":"848-861"},"PeriodicalIF":3.0,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142810544","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Personalized federated learning (PFL) combines client needs and data characteristics to train personalized models for local clients. However, the most of previous PFL schemes encountered challenges such as low model prediction accuracy and privacy leakage when applied to practical datasets. Besides, the existing privacy protection methods fail to achieve satisfactory results in terms of model prediction accuracy and security simultaneously. In this paper, we propose a Privacy-preserving Personalized Federated Learning under Secure Multi-party Computation (SMC-PPFL), which can preserve privacy while obtaining a local personalized model with high prediction accuracy. In SMC-PPFL, noise perturbation is utilized to protect similarity computation, and secure multi-party computation is employed for model sub-aggregations. This combination ensures that clients’ privacy is preserved, and the computed values remain unbiased without compromising security. Then, we propose a weighted sub-aggregation strategy based on the similarity of clients and introduce a regularization term in the local training to improve prediction accuracy. Finally, we evaluate the performance of SMC-PPFL on three common datasets. The experimental results show that SMC-PPFL achieves $2%!sim! 15%$ higher prediction accuracy compared to the previous PFL schemes. Besides, the security analysis also verifies that SMC-PPFL can resist model inversion attacks and membership inference attacks.
{"title":"Secure and Accurate Personalized Federated Learning With Similarity-Based Model Aggregation","authors":"Zhouyong Tan;Junqing Le;Fan Yang;Min Huang;Tao Xiang;Xiaofeng Liao","doi":"10.1109/TSUSC.2024.3403427","DOIUrl":"https://doi.org/10.1109/TSUSC.2024.3403427","url":null,"abstract":"Personalized federated learning (PFL) combines client needs and data characteristics to train personalized models for local clients. However, the most of previous PFL schemes encountered challenges such as low model prediction accuracy and privacy leakage when applied to practical datasets. Besides, the existing privacy protection methods fail to achieve satisfactory results in terms of model prediction accuracy and security simultaneously. In this paper, we propose a <u>P</u>rivacy-preserving <u>P</u>ersonalized <u>F</u>ederated <u>L</u>earning under <u>S</u>ecure <u>M</u>ulti-party <u>C</u>omputation (SMC-PPFL), which can preserve privacy while obtaining a local personalized model with high prediction accuracy. In SMC-PPFL, noise perturbation is utilized to protect similarity computation, and secure multi-party computation is employed for model sub-aggregations. This combination ensures that clients’ privacy is preserved, and the computed values remain unbiased without compromising security. Then, we propose a weighted sub-aggregation strategy based on the similarity of clients and introduce a regularization term in the local training to improve prediction accuracy. Finally, we evaluate the performance of SMC-PPFL on three common datasets. The experimental results show that SMC-PPFL achieves <inline-formula><tex-math>$2%!sim! 15%$</tex-math></inline-formula> higher prediction accuracy compared to the previous PFL schemes. Besides, the security analysis also verifies that SMC-PPFL can resist model inversion attacks and membership inference attacks.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"10 1","pages":"132-145"},"PeriodicalIF":3.0,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143184036","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：