Web Services – ICWS 2019

Social influence is the behavioral change of a person because of the perceived relationship with other people, organizations and society in general. With the exponential growth of online social network services especially mobile messaging apps, users around the world are logging in to messaging apps to not only chat with friends but also to connect with brands, browse merchandise, and watch content. Mobile chat apps boast a number of distinct characteristics that make their audiences particularly appealing to businesses and marketers, including their size, retention and usage rates, and user demographics. The combined user base of the top four chat apps is larger than the combined user base of the top four social networks. Therefore, it makes great sense to analyze user behavior and social influence in mobile messaging apps. In this paper, we focus on computational aspects of measuring social influence of groups formed in mobile messaging apps. We describe the special features of mobile messaging apps and present challenges. We address the challenges by proposing a temporal weighted data model to measure the group influence in messaging apps by considering their special features, with implementation and evaluation in the end.

This paper addresses the pricing problem of a digital services marketplace under asymmetric information. An example is an online learning platform such as Coursera that provides courses from service providers (in this case, universities) to learners. We focus on the matching of digital services to the consumers of these services using partially-observable consumer and service attributes. We develop the optimal pricing policies of the marketplace and show that when the distributions of unobservable valuations are exponential, the marketplace sets a single matching fee (avoiding price-discrimination across providers) which is levied on the less price-sensitive side of the marketplace.

Big data analytics typically requires large amounts of resources to process ever-increasing data volumes. This can be time consuming and result in considerable expenses. Analytics-as-a-Service (AaaS) platforms provide a way to tackle expensive resource costs and lengthy data processing times by leveraging automatic resource management with a pay-per-use service delivery model. This paper explores optimization of resource management algorithms for AaaS platforms to automatically and elastically provision cloud resources to execute queries with Service Level Agreement (SLA) guarantees. We present admission control and cloud resource scheduling algorithms that serve multiple objectives including profit maximization for AaaS platform providers and query time minimization for users. Moreover, to enable queries that require timely responses and/or have constrained budgets, we apply data sampling-based admission control and resource scheduling where accuracy can be traded-off for reduced costs and quicker responses when necessary. We conduct extensive experimental evaluations for the algorithm performances compared to state-of-the-art algorithms. Experiment results show that our proposed algorithms perform significantly better in increasing query admission rates, consuming less resources and hence reducing costs, and ultimately provide a more flexible resource management solution for fast, cost-effective, and reliable big data processing.

Web service composition adopts functional features including the inputs and outputs, and non-functional features including quality of service (QoS), conditional structure constraints, user preferences, and trusts to compose homogeneous or heterogeneous services together in order to create value-added services. However, in some complex practical application scenarios, the web services with the same function can provide the generous differentiated contents, and there is no approach to focus on the user’s constraints on the content provided by the web services. In this paper, we focus on handling three composition dimensions simultaneously including functional features, QoS and the user’s constraints on the contents provided by the web services. Therefore, an improved genetic algorithm to obtain an optimal solution for this task is applied. In addition, we also take it into consideration that the over-constrained problem caused by implicit conflicting constraints and improve a constraint correction approach to solve this problem with less cost of consistency checks. Experimental results using the real datasets about travel demonstrate the effectiveness of our approach in creating the fully functional and quality-optimized solutions, on the premise that the users constraints on the content are satisfied.

Anomaly detection is a key challenge in data mining, which refers to finding patterns in data that do not conform to expected behavior. It has a wide range of applications in many fields as diverse as finance, medicine, industry, and the Internet. In particular, intelligent operation has made great progress in recent years and has an urgent need for this technology. In this paper, we study the problem of anomaly detection in the context of intelligent operation and find the practical need for high-accuracy, online and universal anomaly detection algorithms in time series database. Based on the existing algorithms, we propose an innovative online distance-based anomaly detection algorithm. K-means and time-space trade-off mechanism are used to reduce the time complexity. Through the experiments on Yahoo! Web-scope S5 dataset we show that our algorithm can detect anomalies accurately. The comparative study of several anomaly detectors verifies the effectiveness and generality of the proposed algorithm.

Electronic health record (EHR) greatly enhances the convenience of cross-domain sharing and has been proven effectively to improve the quality of healthcare. On the other hand, the sharing of sensitive medical data is facing critical security and privacy issues, which become an obstacle that prevents EHR being widely adopted. In this paper, we address several challenges in very important patients’ (VIPs) data privacy, including how to protect a VIP’s identity by using pseudonym, how to enable a doctor to update an encrypted EHR with the VIP’s absence, how to help a doctor link up and decrypt historical EHRs of a patient for secondary use under a secure environment, and so on. Then we propose a framework for secure EHR data management. In our framework, we use a transitive pseudonym generation technique to allow a patient to vary his/her identity in each hospital visit. We separate metadata from detailed EHR data in storage, so that the security of EHR data is guaranteed by the security of both the central server and local servers in all involved hospitals. Furthermore, in our framework, a hospital can encrypt and upload a patient’s EHR when he/she is absent; a patient can help to download and decrypt his/her previous EHRs from the central server; and a doctor can decrypt a patient’s historical EHRs for secondary use under the help and audit by several proxies.

Server consolidation technique plays an important role in energy management and load-balancing of cloud computing systems. Dynamic virtual machine (VM) consolidation is a promising consolidation approach in this direction, which aims at using least active physical machines (PMs) through appropriately migrating VMs to reduce resource consumption. The resulting optimization problem is well-acknowledged to be NP-hard optimization problems. In this paper, we propose a novel merge-and-split-based coalitional game-theoretic approach for VM consolidation in heterogeneous clouds. The proposed approach first partitions PMs into different groups based on their load levels, then employs a coalitional-game-based VM consolidation algorithm (CGMS) in choosing members from such groups to form effective coalitions, performs VM migrations among the coalition members to maximize the payoff of every coalition, and close PMs with low energy-efficiency. Experimental results based on multiple cases clearly demonstrate that our proposed approach outperforms traditional ones in terms of energy-saving and level of load fairness.

Blockchain has become a hot topic in recent years. Many applications apply permissioned blockchain to achieve secure data sharing across organizations such as healthcare blockchain. In the permissioned blockchain, on the one hand, the blockchain system is required to support efficient and dynamic authentication for adding and deleting users in a distributed environment. On the other hand, in some particular applications such as healthcare domain, users prefer to keep anonymity in the process of authentication. Although many solutions for anonymous authentication have been proposed, they often require the participation of a central trusted party in the process of authentication and are not efficient enough. In this paper, we focus on designing an efficient traceable and anonymous authentication scheme, which supports efficient authentication while without revealing user’s identity information and does not requires the participation of a central trusted party. While, in case of dispute, the identity of users can be revealed. Moreover, the proposed scheme is able to support dynamic adding and deleting users. Finally, we analyze the security and privacy properties of the proposed scheme and evaluate its performance in terms of computational cost. The experimental results show that the proposed scheme is more efficient than exist schemes and can be easily deployed in the permissioned blockchain.

In recent years, the interest in the Internet of Things has been growing, and WSN is a promising technology that could be applied in many situations. Regardless of the nature of the application, WSNs are often used for data acquisition, to obtain information from an environment of interest, so it is essential to consider how this data will be made available to users. Over the last years, an increasing number of web services have been used to deal with databases and final users, providing familiar interfaces and multi-platform access to those data. To address this problem, our study proposes a web application based on MVC architecture to monitor, organize and manage devices and data in a wireless sensor network. Here, is presented a functional evaluation of the proposed system and a discussion regarding the test results.

Performance optimizations of large scale services can lead to significant wins on service efficiency and performance. CPU resource is one of the most common performance bottlenecks, hence improving CPU performance has been the focus of many performance optimization efforts. In particular, reducing iTLB (instruction TLB) miss rates can greatly improve CPU performance and speed up service running.

At Facebook, we have achieved CPU reduction by applying a solution that firstly identifies hot-text of the (software) binary and then places the binary on huge pages (i.e., 2 MB+ memory pages). The solution is wrapped into an automated framework, enabling service owners to effortlessly adopt it. Our framework has been applied to many services at Facebook, and this paper shares our experiences and findings.

AI, machine learning, and deep learning tools have now become easily accessible on the cloud. However, the adoption of these cloud-based services for heavy industries has been limited due to the gap between general purpose AI tools and operational requirements for production industries. There are three fundamentals gaps. The first is the lack of purpose built solution pipelines designed for common industrial problem types, the second is the lack of tools for automating the learning from noisy sensor data and the third is the lack of platforms which help practitioners leverage cloud-based environment for building and deploying custom modeling pipelines. In this paper, we present ThunderML, a toolkit that addresses these gaps by providing powerful programming model that allows rapid authoring, training and deployment for Industry 4.0 applications. Importantly, the system also facilitates cloud-based deployments by providing a vendor agnostic pipeline execution and deployment layer.

The present study examines Twitter conversations around cardiovascular health in order to assess the topical foci of these conversations as well as the role of various state departments of health. After scraping tweets containing relevant keywords, Latent Dirichlet Allocation (LDA) was used to identify the most important topics discussed around the issue, while PageRank was used to determine the relative prominence of different users. The results indicate that a small number of state departments of health play an especially significant role in these conversations. Furthermore, irregular events like ebola outbreaks also exert a strong influence over the volume of tweets made in general by state departments of health.