Deep Reinforcement Learning (DRL) combines the perceptual capabilities of deep learning with the decision-making capabilities of Reinforcement Learning RL, which can achieve enhanced decision-making. However, the environmental state data contains the privacy of the users. There exists consequently a potential risk of environmental state information being leaked during RL training. Some data desensitization and anonymization technologies are currently being used to protect data privacy. There may still be a risk of privacy disclosure with these desensitization techniques. Meanwhile, policymakers need the environmental state to make decisions, which will cause the disclosure of raw environmental data. To address the privacy issues in DRL, we propose a differential privacy-based online DRL algorithm. The algorithm will add Gaussian noise to the gradients of the deep network according to the privacy budget. More important, we prove tighter bounds for the privacy budget. Furthermore, we train an autocoder to protect the raw environmental state data. In this work, we prove the privacy budget formulation for differential privacy-based online deep RL. Experiments show that the proposed algorithm can improve privacy protection while still having relatively excellent decisionmaking performance.
{"title":"A Privacy-Preserving Online Deep Learning Algorithm Based on Differential Privacy","authors":"Jun Li, Fengshi Zhang, Yonghe Guo, Siyuan Li, Guanjun Wu, Dahui Li, Hongsong Zhu","doi":"10.1109/CSCWD57460.2023.10152847","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152847","url":null,"abstract":"Deep Reinforcement Learning (DRL) combines the perceptual capabilities of deep learning with the decision-making capabilities of Reinforcement Learning RL, which can achieve enhanced decision-making. However, the environmental state data contains the privacy of the users. There exists consequently a potential risk of environmental state information being leaked during RL training. Some data desensitization and anonymization technologies are currently being used to protect data privacy. There may still be a risk of privacy disclosure with these desensitization techniques. Meanwhile, policymakers need the environmental state to make decisions, which will cause the disclosure of raw environmental data. To address the privacy issues in DRL, we propose a differential privacy-based online DRL algorithm. The algorithm will add Gaussian noise to the gradients of the deep network according to the privacy budget. More important, we prove tighter bounds for the privacy budget. Furthermore, we train an autocoder to protect the raw environmental state data. In this work, we prove the privacy budget formulation for differential privacy-based online deep RL. Experiments show that the proposed algorithm can improve privacy protection while still having relatively excellent decisionmaking performance.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"29 1","pages":"559-564"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81634318","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152789
Xiang Cao, Ningjiang Chen, Xuemei Yuan, Yifei Liu
With the support of 5G technology, mobile edge computing has made the application of industrial IoT and power IoT more and more extensive. By deploying a certain number of edge servers at the edge of the network, network service delay may significantly reduce. For the IoT scenario where the content demand is unpredictable, there are multiple distributed cloud servers and the distributed cloud servers do not communicate directly, a feasible way to improve the network service quality is to dynamically optimize the storage of edge servers and formulate targeted caching strategies. This paper proposes an edge caching approach based on multi-agent deep deterministic policy gradient named MADDPG-C, which regards distributed cloud servers and edge servers as different types of agents and maximizes the efficiency of edge caching in cooperation and competition. Simulation experiments show that the proposed MADDPG-C can further improve the hit rate of the edge cache and reduce the waiting delay of terminal devices.
{"title":"A Cooperative Edge Caching Approach Based on Multi-Agent Deep Reinforcement Learning","authors":"Xiang Cao, Ningjiang Chen, Xuemei Yuan, Yifei Liu","doi":"10.1109/CSCWD57460.2023.10152789","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152789","url":null,"abstract":"With the support of 5G technology, mobile edge computing has made the application of industrial IoT and power IoT more and more extensive. By deploying a certain number of edge servers at the edge of the network, network service delay may significantly reduce. For the IoT scenario where the content demand is unpredictable, there are multiple distributed cloud servers and the distributed cloud servers do not communicate directly, a feasible way to improve the network service quality is to dynamically optimize the storage of edge servers and formulate targeted caching strategies. This paper proposes an edge caching approach based on multi-agent deep deterministic policy gradient named MADDPG-C, which regards distributed cloud servers and edge servers as different types of agents and maximizes the efficiency of edge caching in cooperation and competition. Simulation experiments show that the proposed MADDPG-C can further improve the hit rate of the edge cache and reduce the waiting delay of terminal devices.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"41 7","pages":"1772-1777"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72471002","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152746
Fengyuan Shi, Zhou-yu Zhou, Wei Yang, Shu Li, Qingyun Liu, Xiuguo Bao
In a static network, attackers can easily launch network attacks on target hosts which have long-term constant IP addresses. In order to defend against attackers effectively, many defense approaches use IP hopping to dynamically transform IP configuration. However, these approaches usually focus on one type of network attacks, scanning attacks or Denial of Service (DoS) attacks, and cannot sense network situations. This paper proposes AHIP, an adaptive IP hopping method for moving target defense (MTD) to defend against different network attacks. We use a trained lightweight one-dimensional convolutional neural network (1D-CNN) detector to judge whether there are no attacks, scanning attacks or DoS attacks in the network, which can adaptively trigger corresponding IP hopping strategy. We use specific hardware and software to create the software defined network (SDN) environment for experiments. The experiments prove that AHIP performs better to thwart network attacks and has lower system overhead.
{"title":"AHIP: An Adaptive IP Hopping Method for Moving Target Defense to Thwart Network Attacks","authors":"Fengyuan Shi, Zhou-yu Zhou, Wei Yang, Shu Li, Qingyun Liu, Xiuguo Bao","doi":"10.1109/CSCWD57460.2023.10152746","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152746","url":null,"abstract":"In a static network, attackers can easily launch network attacks on target hosts which have long-term constant IP addresses. In order to defend against attackers effectively, many defense approaches use IP hopping to dynamically transform IP configuration. However, these approaches usually focus on one type of network attacks, scanning attacks or Denial of Service (DoS) attacks, and cannot sense network situations. This paper proposes AHIP, an adaptive IP hopping method for moving target defense (MTD) to defend against different network attacks. We use a trained lightweight one-dimensional convolutional neural network (1D-CNN) detector to judge whether there are no attacks, scanning attacks or DoS attacks in the network, which can adaptively trigger corresponding IP hopping strategy. We use specific hardware and software to create the software defined network (SDN) environment for experiments. The experiments prove that AHIP performs better to thwart network attacks and has lower system overhead.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"53 1","pages":"1300-1305"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76483630","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152708
Yi Gao, Fangfang Yuan, Cong Cao, Majing Su, Dakui Wang, Yanbing Liu
The Domain Name System (DNS), one of the essential basic services on the Internet, is often abused by attackers to launch various cyber attacks, such as phishing and spamming. Researchers have proposed many machine learning-based and deep learning-based methods to detect malicious domains. However, these methods rely on a large-scale dataset with labeled samples for model training. The fact is that the labeled domain samples are limited in the real-world DNS dataset. In this paper, we propose a few-shot malicious domain detection model named MetaDom, which employs a meta-learning algorithm for model optimization. Specifically, We first model the DNS scenario as a heterogeneous graph to capture richer information by analysing the complex relations among domains, IP addresses and clients. Then, we learn the domain representations with a heterogeneous graph neural network on the DNS HG. Finally, considering that only few labeled data are available in the real-world DNS scenario, a meta-learning algorithm with knowledge distillation is introduced to optimize the model. Extensive experiments on the real DNS dataset show that MetaDom outperforms other state-of-the-art methods.
{"title":"Few-shot Malicious Domain Detection on Heterogeneous Graph with Meta-learning","authors":"Yi Gao, Fangfang Yuan, Cong Cao, Majing Su, Dakui Wang, Yanbing Liu","doi":"10.1109/CSCWD57460.2023.10152708","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152708","url":null,"abstract":"The Domain Name System (DNS), one of the essential basic services on the Internet, is often abused by attackers to launch various cyber attacks, such as phishing and spamming. Researchers have proposed many machine learning-based and deep learning-based methods to detect malicious domains. However, these methods rely on a large-scale dataset with labeled samples for model training. The fact is that the labeled domain samples are limited in the real-world DNS dataset. In this paper, we propose a few-shot malicious domain detection model named MetaDom, which employs a meta-learning algorithm for model optimization. Specifically, We first model the DNS scenario as a heterogeneous graph to capture richer information by analysing the complex relations among domains, IP addresses and clients. Then, we learn the domain representations with a heterogeneous graph neural network on the DNS HG. Finally, considering that only few labeled data are available in the real-world DNS scenario, a meta-learning algorithm with knowledge distillation is introduced to optimize the model. Extensive experiments on the real DNS dataset show that MetaDom outperforms other state-of-the-art methods.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"62 1","pages":"727-732"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76791528","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152792
Fuqing Zhao, Yuebao Liu, Tianpeng Xu
Selecting appropriate algorithms for specific problems has become a significant challenge with the remarkable growth of heuristics and meta-heuristics. To address this challenge, an iterative greedy selection hyper-heuristic algorithm with linear population size reduction (LIGSHH) was proposed in this paper. Using an iterative greedy strategy to choose the high level of exploration, this heuristic selects the Low-Level Heuristics (LLHs) that best suit the current problem. Nine LLHs are specifically designed for continuous optimization problems. Additionally, the exploration and exploitation capabilities of the LIGSHH are balanced by reducing the population size linearly at different stages of the problem. The proposed LIGSHH algorithm and comparison algorithms are tested on the CEC2017 benchmark test suite, and the experimental results show that the LIGSHH algorithm outperforms other comparison algorithms.
{"title":"Iterative Greedy Selection Hyper-heuristic with Linear Population Size Reduction","authors":"Fuqing Zhao, Yuebao Liu, Tianpeng Xu","doi":"10.1109/CSCWD57460.2023.10152792","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152792","url":null,"abstract":"Selecting appropriate algorithms for specific problems has become a significant challenge with the remarkable growth of heuristics and meta-heuristics. To address this challenge, an iterative greedy selection hyper-heuristic algorithm with linear population size reduction (LIGSHH) was proposed in this paper. Using an iterative greedy strategy to choose the high level of exploration, this heuristic selects the Low-Level Heuristics (LLHs) that best suit the current problem. Nine LLHs are specifically designed for continuous optimization problems. Additionally, the exploration and exploitation capabilities of the LIGSHH are balanced by reducing the population size linearly at different stages of the problem. The proposed LIGSHH algorithm and comparison algorithms are tested on the CEC2017 benchmark test suite, and the experimental results show that the LIGSHH algorithm outperforms other comparison algorithms.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"151 1","pages":"751-755"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79549566","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 : 2023-05-24DOI: 10.1109/cscwd57460.2023.10152801
{"title":"Keynote 2 : Promoting the diversity of digital technologies","authors":"","doi":"10.1109/cscwd57460.2023.10152801","DOIUrl":"https://doi.org/10.1109/cscwd57460.2023.10152801","url":null,"abstract":"","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"174 1","pages":""},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79664019","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152680
Xiaobo Wei, Peng Li, Weiyi Huang, Zhiyuan Liu, Qin Liu
Ridesharing benefits the economy and the environment. In multi-hop ridesharing, passengers are permitted to switch vehicles within a single trip, extending the flexibility of conventional ridesharing. Nonetheless, vehicle dispatch is a difficult issue in multi-hop ridesharing. We subdivide the vehicle dispatching problem into the vehicle pairing problem and the request selection problem within a vehicle pair. To address these subproblems, we propose a two-stage framework for vehicle pair dispatching. In the initial stage, we model the vehicle pairing problem as a maximum vehicle-vehicle matching problem in a general graph, which differs from the conventional vehicle-request matching problem in a bipartite graph. The vehicle pairing algorithm is proposed to efficiently solve the vehicle pairing problem. In the second stage, we model the request selection problem as a multidimensional knapsack problem (d-KP) and propose an LP-relaxation request selection algorithm with an approximation ratio 1/5. Experiments conducted on a real-world dataset demonstrate the economic benefit of our proposed two-stage framework.
{"title":"Two-stage Vehicle Pair Dispatch in Multi-hop Ridesharing","authors":"Xiaobo Wei, Peng Li, Weiyi Huang, Zhiyuan Liu, Qin Liu","doi":"10.1109/CSCWD57460.2023.10152680","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152680","url":null,"abstract":"Ridesharing benefits the economy and the environment. In multi-hop ridesharing, passengers are permitted to switch vehicles within a single trip, extending the flexibility of conventional ridesharing. Nonetheless, vehicle dispatch is a difficult issue in multi-hop ridesharing. We subdivide the vehicle dispatching problem into the vehicle pairing problem and the request selection problem within a vehicle pair. To address these subproblems, we propose a two-stage framework for vehicle pair dispatching. In the initial stage, we model the vehicle pairing problem as a maximum vehicle-vehicle matching problem in a general graph, which differs from the conventional vehicle-request matching problem in a bipartite graph. The vehicle pairing algorithm is proposed to efficiently solve the vehicle pairing problem. In the second stage, we model the request selection problem as a multidimensional knapsack problem (d-KP) and propose an LP-relaxation request selection algorithm with an approximation ratio 1/5. Experiments conducted on a real-world dataset demonstrate the economic benefit of our proposed two-stage framework.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"67 1","pages":"255-260"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80437028","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152645
Chuang Zhang, Geng Sun, Jiahui Li, Xiaoya Zheng
Unmanned aerial vehicles (UAVs) as the aerial relay become a highly desired scheme to assist terrestrial network. In this work, we intend to utilize the UAV swarm to assist the communication between the base station (BS) equipped with the planar array antenna (PAA) and the IoT devices by collaborative beamforming (CB). Specifically, we formulate an average achievable rate and energy bi-objective optimization problem (AREBOP) to improve the average achievable rate of IoT terminal devices and energy consumption of UAV swarm by jointly optimize the excitation current weights of BS and UAVs, the position of UAVs and user association order of IoT terminal devices. Moreover, the formulated AREBOP is proved to be NP-hard. Thus, we proposed an multi-objective grasshopper algorithm with specific initialization (MOGOASI) to solve this problem. Simulation results show the effectiveness of MOGOASI and illustrate that the performance of MOGOASI is superior compared to some benchmarks.
{"title":"Bi-objective Optimization for UAV Swarm-enabled Relay Communications via Collaborative Beamforming","authors":"Chuang Zhang, Geng Sun, Jiahui Li, Xiaoya Zheng","doi":"10.1109/CSCWD57460.2023.10152645","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152645","url":null,"abstract":"Unmanned aerial vehicles (UAVs) as the aerial relay become a highly desired scheme to assist terrestrial network. In this work, we intend to utilize the UAV swarm to assist the communication between the base station (BS) equipped with the planar array antenna (PAA) and the IoT devices by collaborative beamforming (CB). Specifically, we formulate an average achievable rate and energy bi-objective optimization problem (AREBOP) to improve the average achievable rate of IoT terminal devices and energy consumption of UAV swarm by jointly optimize the excitation current weights of BS and UAVs, the position of UAVs and user association order of IoT terminal devices. Moreover, the formulated AREBOP is proved to be NP-hard. Thus, we proposed an multi-objective grasshopper algorithm with specific initialization (MOGOASI) to solve this problem. Simulation results show the effectiveness of MOGOASI and illustrate that the performance of MOGOASI is superior compared to some benchmarks.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"111 1","pages":"984-989"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80575002","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152738
Lifu Wang, K. Dong, Xiaodan Gu, Zhen Ling, Ming Yang
IFTTT is one of the most popular Trigger-Action Programming platforms. The rules generated in IFTTT are named IoT Applets. Despite the powerful programming interface provided by IFTTT, establishing an Applet requires technical skills and is not convenient enough for most users. To address this problem, we propose a gesture based programming method to help end users establish and manage IoT Applets in a convenient way. It requires employment of an RGB-D camera, and recognizes users’ pointing rays and hand actions. The obtained information is interpreted to certain devices and device events for Applet management. An experiment involving 20 participants validates the performance of our proposed method.
{"title":"Lightweight Gesture Based Trigger-Action Programming for Home Internet-of-Things","authors":"Lifu Wang, K. Dong, Xiaodan Gu, Zhen Ling, Ming Yang","doi":"10.1109/CSCWD57460.2023.10152738","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152738","url":null,"abstract":"IFTTT is one of the most popular Trigger-Action Programming platforms. The rules generated in IFTTT are named IoT Applets. Despite the powerful programming interface provided by IFTTT, establishing an Applet requires technical skills and is not convenient enough for most users. To address this problem, we propose a gesture based programming method to help end users establish and manage IoT Applets in a convenient way. It requires employment of an RGB-D camera, and recognizes users’ pointing rays and hand actions. The obtained information is interpreted to certain devices and device events for Applet management. An experiment involving 20 participants validates the performance of our proposed method.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"57 1","pages":"959-964"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80865892","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 : 2023-05-24DOI: 10.1109/CSCWD57460.2023.10152842
Wangbing Cheng, MinFeng Zhang, Fang Dong, Shucun Fu
Multi-view inference can utilize visual information from several views like a human being and significantly improve accuracy in some scenes, but it inevitably incurs more computing overhead than traditional DNN inference. To meet the requirement of low latency in typical scenarios, we consider utilizing model partition technique of edge computing to speed up multi-view inference, and design a multi-view end-edge co-inference execution framework (MV-IEF) which can make use of both end and edge resources for multi-view inference tasks. However, when employing the framework simply, the efficiency of multi-view inference will be constrained by network dynamics and heterogeneity of devices corresponding to multiple views. To break this constraint, we establish an optimization model based on the framework to minimize the multi-view inference time and solve it on the basis of game theory. And meanwhile, we propose a joint optimization algorithm for multi-view resource allocation and model partition (MV-JRAMP), which can make remarkable decisions of resource allocation and model partiton according to network status and computing capabilities of devices. Finally, we build a prototype and evaluate the performance of MV-JRAMP. Experiments show that MV-JRAMP can accelerate multi-view inference by up to 3.71×.
{"title":"Accelerate Multi-view Inference with End-edge Collaborative Computing","authors":"Wangbing Cheng, MinFeng Zhang, Fang Dong, Shucun Fu","doi":"10.1109/CSCWD57460.2023.10152842","DOIUrl":"https://doi.org/10.1109/CSCWD57460.2023.10152842","url":null,"abstract":"Multi-view inference can utilize visual information from several views like a human being and significantly improve accuracy in some scenes, but it inevitably incurs more computing overhead than traditional DNN inference. To meet the requirement of low latency in typical scenarios, we consider utilizing model partition technique of edge computing to speed up multi-view inference, and design a multi-view end-edge co-inference execution framework (MV-IEF) which can make use of both end and edge resources for multi-view inference tasks. However, when employing the framework simply, the efficiency of multi-view inference will be constrained by network dynamics and heterogeneity of devices corresponding to multiple views. To break this constraint, we establish an optimization model based on the framework to minimize the multi-view inference time and solve it on the basis of game theory. And meanwhile, we propose a joint optimization algorithm for multi-view resource allocation and model partition (MV-JRAMP), which can make remarkable decisions of resource allocation and model partiton according to network status and computing capabilities of devices. Finally, we build a prototype and evaluate the performance of MV-JRAMP. Experiments show that MV-JRAMP can accelerate multi-view inference by up to 3.71×.","PeriodicalId":51008,"journal":{"name":"Computer Supported Cooperative Work-The Journal of Collaborative Computing","volume":"42 1","pages":"1625-1631"},"PeriodicalIF":2.4,"publicationDate":"2023-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80875422","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：