Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00046
Benjamin Weder, Uwe Breitenbücher, F. Leymann, Karoline Wild
Quantum computing has the potential to significantly impact many application domains, as several quantum algorithms are promising to solve problems more efficiently than possible on classical computers. However, various complex pre- and post-processing tasks have to be performed when executing a quantum circuit, which require immense mathematical and technical knowledge. For example, calculations on today’s quantum computers are noisy and require an error mitigation task after the execution. Hence, integrating classical applications with quantum circuits is a difficult challenge. In this paper, we introduce a modeling extension for imperative workflow languages to enable the integration of quantum computations and ease the orchestration of classical applications and quantum circuits. Further, we show how the extension can be mapped to native modeling constructs of extended workflow languages to retain the portability of the workflows. We validate the practical feasibility of our approach by applying our proposed extension to BPMN and introduce Quantum4BPMN.
{"title":"Integrating Quantum Computing into Workflow Modeling and Execution","authors":"Benjamin Weder, Uwe Breitenbücher, F. Leymann, Karoline Wild","doi":"10.1109/UCC48980.2020.00046","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00046","url":null,"abstract":"Quantum computing has the potential to significantly impact many application domains, as several quantum algorithms are promising to solve problems more efficiently than possible on classical computers. However, various complex pre- and post-processing tasks have to be performed when executing a quantum circuit, which require immense mathematical and technical knowledge. For example, calculations on today’s quantum computers are noisy and require an error mitigation task after the execution. Hence, integrating classical applications with quantum circuits is a difficult challenge. In this paper, we introduce a modeling extension for imperative workflow languages to enable the integration of quantum computations and ease the orchestration of classical applications and quantum circuits. Further, we show how the extension can be mapped to native modeling constructs of extended workflow languages to retain the portability of the workflows. We validate the practical feasibility of our approach by applying our proposed extension to BPMN and introduce Quantum4BPMN.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130953306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00038
Victor Millnert, Johan Eker
Elastic and scalable compute resources are a fundamental part of cloud computing. Efficient management of cloud resources is crucial in order to provide high quality services and applications. In this work we present a novel method for scaling cloud resources and provide stability guarantees. We do this by leveraging ideas and concepts from classic control theory, namely mid-range control and combine horizontal scaling and vertical scaling in a novel way. Horizontal scaling is typically when one adds/removes whole unites of resources (e.g., virtual machines or containers), while vertical scaling is when one grows/shrinks already allocated resources (e.g., making a deployed virtual machine larger/smaller). Each methods has their own trade-offs: i) horizontal scaling is often slow and coarse-grained, but can scale over a large range, and ii) vertical scaling is often quick and smooth, but has limited range.The proposed algorithm is called HoloScale, which leverages the strengths of both scaling mechanisms, without the drawbacks. The method is capable of scaling smoothly, quickly, and over a large range. By using core concepts from control theory, we show that systems managed by the HoloScale algorithm are stable in the presence of time-varying scaling delays.
{"title":"HoloScale: horizontal and vertical scaling of cloud resources","authors":"Victor Millnert, Johan Eker","doi":"10.1109/UCC48980.2020.00038","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00038","url":null,"abstract":"Elastic and scalable compute resources are a fundamental part of cloud computing. Efficient management of cloud resources is crucial in order to provide high quality services and applications. In this work we present a novel method for scaling cloud resources and provide stability guarantees. We do this by leveraging ideas and concepts from classic control theory, namely mid-range control and combine horizontal scaling and vertical scaling in a novel way. Horizontal scaling is typically when one adds/removes whole unites of resources (e.g., virtual machines or containers), while vertical scaling is when one grows/shrinks already allocated resources (e.g., making a deployed virtual machine larger/smaller). Each methods has their own trade-offs: i) horizontal scaling is often slow and coarse-grained, but can scale over a large range, and ii) vertical scaling is often quick and smooth, but has limited range.The proposed algorithm is called HoloScale, which leverages the strengths of both scaling mechanisms, without the drawbacks. The method is capable of scaling smoothly, quickly, and over a large range. By using core concepts from control theory, we show that systems managed by the HoloScale algorithm are stable in the presence of time-varying scaling delays.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129312626","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00051
Y. Yan
As people keep a watch eye on the oral health, more people choose to go to professional dental hospitals for the regular dental examinations and diagnosis. It is well known that the dental diagnosis and treatment require excellent nursing skills and extensive experience by the dentists. Nervously, the number of experts is limited. However, the rapid increase in the number of diagnoses and the small number of professional dentists resulted in an increase in the daily diagnostic frequency of dentists, and the overworked working hours seriously affected the energy and diagnostic efficiency of dentists. This study for the sake of reduce the burden of dental diagnosis, proposes a computer-aided diagnosis method. This method classifies gingivitis images by using the image feature extraction method of fractional Fourier entropy (FRFE) and biogeography-based optimization (BBO) algorithm. The FRFE coefficient extracted from the image was used as the input feature vector, and the classification was carried out by the BBO algorithm with the optimal scheme of automatic screening. After 10-fold cross-validation, more effective healthy and pathological gingival image classification results were obtained compared with the latest methods.
{"title":"Gingivitis detection by Fractional Fourier Entropy and Biogeography-based Optimization","authors":"Y. Yan","doi":"10.1109/UCC48980.2020.00051","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00051","url":null,"abstract":"As people keep a watch eye on the oral health, more people choose to go to professional dental hospitals for the regular dental examinations and diagnosis. It is well known that the dental diagnosis and treatment require excellent nursing skills and extensive experience by the dentists. Nervously, the number of experts is limited. However, the rapid increase in the number of diagnoses and the small number of professional dentists resulted in an increase in the daily diagnostic frequency of dentists, and the overworked working hours seriously affected the energy and diagnostic efficiency of dentists. This study for the sake of reduce the burden of dental diagnosis, proposes a computer-aided diagnosis method. This method classifies gingivitis images by using the image feature extraction method of fractional Fourier entropy (FRFE) and biogeography-based optimization (BBO) algorithm. The FRFE coefficient extracted from the image was used as the input feature vector, and the classification was carried out by the BBO algorithm with the optimal scheme of automatic screening. After 10-fold cross-validation, more effective healthy and pathological gingival image classification results were obtained compared with the latest methods.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125139067","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00058
Stoddard Rosenkrantz, Huiyang Li, Prathyusha Enganti, Zhongwei Li, Lin Sun, Zhijun Wang, Hao Che, Hong Jiang
As the IoT-Edge-Cloud hierarchy is evolving into a mature ecosystem, large-scale Sensing-as-a-Service (SaS) based services with stringent job service level objectives (SLOs) are expected to emerge as dominant cloud services. A viable business model for SaS must be inherently multi-tier by design and work in a confederated environment involving a large number of voluntary stakeholders who may appear at different tiers. It must also honor privacy and autonomous control of stakeholder resources. This calls for a fully distributed, SLO-aware job resource allocation and scheduling platform to be developed. In this paper, we propose a tail-latency-SLO-aware job resource allocation and scheduling platform for SaS, called JADE. It is a four-tier platform, i.e., cloud, edge cluster, edge, and IoT tiers. To honor the privacy and autonomy of control for individual stakeholders at different tiers, the JADE design follows the design principle of separation of concerns among tiers. Central to its design is to develop a decomposition technique that decomposes SaS service requirements, in particular, the job tail-latency SLO, into task performance budgets for individual sensing tasks mapped to each lower tier. This makes it possible to allow each lower tier to manage its own resources autonomously to meet the sensing task budgets and hence the SaS service requirements, while preserving its privacy and autonomy of control. Finally, preliminary testing results based on both simulation and an initial prototype of JADE are presented to demonstrate the promising prospects of the solution.
{"title":"JADE: Tail-Latency-SLO-Aware Job Scheduling for Sensing-as-a-Service","authors":"Stoddard Rosenkrantz, Huiyang Li, Prathyusha Enganti, Zhongwei Li, Lin Sun, Zhijun Wang, Hao Che, Hong Jiang","doi":"10.1109/UCC48980.2020.00058","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00058","url":null,"abstract":"As the IoT-Edge-Cloud hierarchy is evolving into a mature ecosystem, large-scale Sensing-as-a-Service (SaS) based services with stringent job service level objectives (SLOs) are expected to emerge as dominant cloud services. A viable business model for SaS must be inherently multi-tier by design and work in a confederated environment involving a large number of voluntary stakeholders who may appear at different tiers. It must also honor privacy and autonomous control of stakeholder resources. This calls for a fully distributed, SLO-aware job resource allocation and scheduling platform to be developed. In this paper, we propose a tail-latency-SLO-aware job resource allocation and scheduling platform for SaS, called JADE. It is a four-tier platform, i.e., cloud, edge cluster, edge, and IoT tiers. To honor the privacy and autonomy of control for individual stakeholders at different tiers, the JADE design follows the design principle of separation of concerns among tiers. Central to its design is to develop a decomposition technique that decomposes SaS service requirements, in particular, the job tail-latency SLO, into task performance budgets for individual sensing tasks mapped to each lower tier. This makes it possible to allow each lower tier to manage its own resources autonomously to meet the sensing task budgets and hence the SaS service requirements, while preserving its privacy and autonomy of control. Finally, preliminary testing results based on both simulation and an initial prototype of JADE are presented to demonstrate the promising prospects of the solution.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127183024","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00065
Yi Han, Chenxi Yu, Dongdong Li, Jie Zhang, Yunqiao Lai
Due to the rapid development of telecommunication technology, the application of 360° video is increasingly gaining attention. For the design of the 360° video transmission mechanism, quality of experience (QoE) from the customer’s perspective is very important. This study can help the readers to understand advantages/limitations of different 360° video quality assessment methods and be able to make suitable choices for various systems. To make it clearer, this paper performs experiments with two steps. Experiment I compares different assessment methods for evaluating 360° video quality selected from online and offline methods, respectively. Experiment II studies the performance of these assessment methods on different video quality levels. The results show that both offline and online test results have a relatively good correlation with the subjective test results. This paper statistically evaluates and compares the accuracy of different 360° video QoE assessment methods and the conclusion drawn from this paper can be used as a guideline when designing adaptive 360° video streaming systems and can also be applied to cloud computing in the future work.
{"title":"Accuracy Analysis on 360° Virtual Reality Video Quality Assessment Methods","authors":"Yi Han, Chenxi Yu, Dongdong Li, Jie Zhang, Yunqiao Lai","doi":"10.1109/UCC48980.2020.00065","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00065","url":null,"abstract":"Due to the rapid development of telecommunication technology, the application of 360° video is increasingly gaining attention. For the design of the 360° video transmission mechanism, quality of experience (QoE) from the customer’s perspective is very important. This study can help the readers to understand advantages/limitations of different 360° video quality assessment methods and be able to make suitable choices for various systems. To make it clearer, this paper performs experiments with two steps. Experiment I compares different assessment methods for evaluating 360° video quality selected from online and offline methods, respectively. Experiment II studies the performance of these assessment methods on different video quality levels. The results show that both offline and online test results have a relatively good correlation with the subjective test results. This paper statistically evaluates and compares the accuracy of different 360° video QoE assessment methods and the conclusion drawn from this paper can be used as a guideline when designing adaptive 360° video streaming systems and can also be applied to cloud computing in the future work.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129780562","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/ucc48980.2020.00003
{"title":"[Copyright notice]","authors":"","doi":"10.1109/ucc48980.2020.00003","DOIUrl":"https://doi.org/10.1109/ucc48980.2020.00003","url":null,"abstract":"","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"4 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124332815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00057
Krittin Intharawijitr, P. Harvey, Pierre Imai
Regardless of the setting, edge computing has drawn much attention from both the academic and industrial communities. For edge computing, content delivery networks are both a concrete and production deployable use case. While viable at the WAN or telco edge scale, it is unclear if this extends to others, such as in home WiFi routers, as has been assumed by some.In this work-in-progress, we present an initial study on the viability of using smart edge WiFi routers as a caching location. We describe the simulator we created to test this, as well as the analysis of the results obtained. We use 1 day of e-commerce web log traffic from a public data set, as well as a sampled subset of our own site - part of an ecosystem of over 111 million users. We show that in the best case scenario, smart edge routers are inappropriate for e-commerce web caching.
{"title":"A Feasibility Study of Cache in Smart Edge Router for Web-Access Accelerator","authors":"Krittin Intharawijitr, P. Harvey, Pierre Imai","doi":"10.1109/UCC48980.2020.00057","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00057","url":null,"abstract":"Regardless of the setting, edge computing has drawn much attention from both the academic and industrial communities. For edge computing, content delivery networks are both a concrete and production deployable use case. While viable at the WAN or telco edge scale, it is unclear if this extends to others, such as in home WiFi routers, as has been assumed by some.In this work-in-progress, we present an initial study on the viability of using smart edge WiFi routers as a caching location. We describe the simulator we created to test this, as well as the analysis of the results obtained. We use 1 day of e-commerce web log traffic from a public data set, as well as a sampled subset of our own site - part of an ecosystem of over 111 million users. We show that in the best case scenario, smart edge routers are inappropriate for e-commerce web caching.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122851416","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00067
Petr Mrozek, John Panneerselvam, O. Bagdasar
The rapid growth of e-commerce and online shopping have resulted in an unprecedented increase in the amount of money that is annually lost to credit card fraudsters. In an attempt to address credit card fraud, researchers are leveraging the application of various machine learning techniques for efficiently detecting and preventing fraudulent credit card transactions. One of the prevalent common issues around the analytics of credit card transactions is the highly unbalanced nature of the datasets, which is frequently associated with the binary classification problems. This paper intends to review, analyse and implement a selection of notable machine learning algorithms such as Logistic Regression, Random Forest, K-Nearest Neighbours and Stochastic Gradient Descent, with the motivation of empirically evaluating their efficiencies in handling unbalanced datasets whilst detecting credit card fraud transactions. A publicly available dataset comprising 284807 transactions of European cardholders is analysed and trained with the studied machine learning techniques to detect fraudulent transactions. Furthermore, this paper also evaluates the incorporation of two notable resampling methods, namely Random Under-sampling and Synthetic Majority Oversampling Techniques (SMOTE) in the aforementioned algorithms, in order to analyse their efficiency in handling unbalanced datasets. The proposed resampling methods significantly increased the detection ability, the most successful technique of combination of Random Forest with Random Under-sampling achieved the recall score of 100% in contrast to the recall score 77% of model without resampling technique. The key contribution of this paper is the postulation of efficient machine learning algorithms together with suitable resampling methods, suitable for credit card fraud detection with unbalanced dataset.
{"title":"Efficient Resampling for Fraud Detection During Anonymised Credit Card Transactions with Unbalanced Datasets","authors":"Petr Mrozek, John Panneerselvam, O. Bagdasar","doi":"10.1109/UCC48980.2020.00067","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00067","url":null,"abstract":"The rapid growth of e-commerce and online shopping have resulted in an unprecedented increase in the amount of money that is annually lost to credit card fraudsters. In an attempt to address credit card fraud, researchers are leveraging the application of various machine learning techniques for efficiently detecting and preventing fraudulent credit card transactions. One of the prevalent common issues around the analytics of credit card transactions is the highly unbalanced nature of the datasets, which is frequently associated with the binary classification problems. This paper intends to review, analyse and implement a selection of notable machine learning algorithms such as Logistic Regression, Random Forest, K-Nearest Neighbours and Stochastic Gradient Descent, with the motivation of empirically evaluating their efficiencies in handling unbalanced datasets whilst detecting credit card fraud transactions. A publicly available dataset comprising 284807 transactions of European cardholders is analysed and trained with the studied machine learning techniques to detect fraudulent transactions. Furthermore, this paper also evaluates the incorporation of two notable resampling methods, namely Random Under-sampling and Synthetic Majority Oversampling Techniques (SMOTE) in the aforementioned algorithms, in order to analyse their efficiency in handling unbalanced datasets. The proposed resampling methods significantly increased the detection ability, the most successful technique of combination of Random Forest with Random Under-sampling achieved the recall score of 100% in contrast to the recall score 77% of model without resampling technique. The key contribution of this paper is the postulation of efficient machine learning algorithms together with suitable resampling methods, suitable for credit card fraud detection with unbalanced dataset.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"71 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117085813","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00069
Taehee Jeong, Ehsam Ghasemi, Jorn Tuyls, Elliott Delaye, Ashish Sirasao
Neural network pruning is a critical technique to efficiently deploy neural network models on edge devices with limited computing resources. Although many neural network pruning methods have been published, it is difficult to implement such algorithms due to their inherent complexity. In this work, we propose a functional pruning tool for neural network models. Our pruning procedure is simple and easy to be implemented, and efficient for deployment. Our pruning tool automatically detects redundancy inside neural network models and prunes the redundant channels. Doing so reduces the total number of model parameters and hence, compresses the size of the model. This approach significantly reduces the number of FLOPs needed for executing the neural network model and improves the inference runtime. To further improve the inference runtime of the pruned model, we leveraged Apache TVM to deploy the pruned model on the DPU FPGA-based hardware accelerator. To demonstrate our approach, we pruned the VGG-16 model on Flower dataset and reached 53-fold reduction in model size with only 7% drop in validation accuracy. The inference latency is reduced 4-fold on CPU and 16-fold on FPGA for the pruned models, compared with the latency of the base model on CPU.
{"title":"Neural network pruning and hardware acceleration","authors":"Taehee Jeong, Ehsam Ghasemi, Jorn Tuyls, Elliott Delaye, Ashish Sirasao","doi":"10.1109/UCC48980.2020.00069","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00069","url":null,"abstract":"Neural network pruning is a critical technique to efficiently deploy neural network models on edge devices with limited computing resources. Although many neural network pruning methods have been published, it is difficult to implement such algorithms due to their inherent complexity. In this work, we propose a functional pruning tool for neural network models. Our pruning procedure is simple and easy to be implemented, and efficient for deployment. Our pruning tool automatically detects redundancy inside neural network models and prunes the redundant channels. Doing so reduces the total number of model parameters and hence, compresses the size of the model. This approach significantly reduces the number of FLOPs needed for executing the neural network model and improves the inference runtime. To further improve the inference runtime of the pruned model, we leveraged Apache TVM to deploy the pruned model on the DPU FPGA-based hardware accelerator. To demonstrate our approach, we pruned the VGG-16 model on Flower dataset and reached 53-fold reduction in model size with only 7% drop in validation accuracy. The inference latency is reduced 4-fold on CPU and 16-fold on FPGA for the pruned models, compared with the latency of the base model on CPU.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123964139","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-01DOI: 10.1109/UCC48980.2020.00035
Louis-Henri Merino, M. Cukier
Preventing and detecting attacks in the cloud are difficult tasks involving technical, financial, and legal challenges. Baseline security solutions from cloud providers are often inadequate to secure cloud instances properly. In addition, entry-level cloud instances offer few resources, as little as 512MB of RAM, and particular actions are either costly or limited by cloud providers, hindering the operation of commercial security solutions, such as antivirus software, and intrusion detection and prevention (IDP) systems. State-of-the-art research IDP systems have made great progress using machine and deep learning but they encounter certain limitations when operating in the cloud. We introduce Xshield, a lightweight IDP framework designed for the cloud, that consists of a limited number of Producers constantly gathering malicious information, analyzing it through one or more arbitrary intrusion detection and/or prevention strategies and passing the processed information along to Consumers, an IDP agent on cloud customers’ instances. We implement and evaluate a Producer prototype by deploying 138 Producers on a cloud provider across 15 regions for seven days and use the collected information to demonstrate how a limited number but strategically placed Producers are capable of protecting cloud customers’ instances as well as present insights on attacker behavior in the cloud. We then discuss, based on attacker behavior insights, what kind of existing IDP strategies can be adapted to operate on Producers.
{"title":"An Approach for Preventing and Detecting Attacks in the Cloud","authors":"Louis-Henri Merino, M. Cukier","doi":"10.1109/UCC48980.2020.00035","DOIUrl":"https://doi.org/10.1109/UCC48980.2020.00035","url":null,"abstract":"Preventing and detecting attacks in the cloud are difficult tasks involving technical, financial, and legal challenges. Baseline security solutions from cloud providers are often inadequate to secure cloud instances properly. In addition, entry-level cloud instances offer few resources, as little as 512MB of RAM, and particular actions are either costly or limited by cloud providers, hindering the operation of commercial security solutions, such as antivirus software, and intrusion detection and prevention (IDP) systems. State-of-the-art research IDP systems have made great progress using machine and deep learning but they encounter certain limitations when operating in the cloud. We introduce Xshield, a lightweight IDP framework designed for the cloud, that consists of a limited number of Producers constantly gathering malicious information, analyzing it through one or more arbitrary intrusion detection and/or prevention strategies and passing the processed information along to Consumers, an IDP agent on cloud customers’ instances. We implement and evaluate a Producer prototype by deploying 138 Producers on a cloud provider across 15 regions for seven days and use the collected information to demonstrate how a limited number but strategically placed Producers are capable of protecting cloud customers’ instances as well as present insights on attacker behavior in the cloud. We then discuss, based on attacker behavior insights, what kind of existing IDP strategies can be adapted to operate on Producers.","PeriodicalId":125849,"journal":{"name":"2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129637887","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","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：