Chuanchang Liu

A cost-aware auto-scaling approach using the workload prediction in service clouds

Service clouds are distributed infrastructures which deploys communication services in clouds. The scalability is an important characteristic of service clouds. With the scalability, the service cloud can offer on-demand computing power and storage capacities to different services. In order to achieve the scalability, we need to know when and how to scale virtual resources assigned to different services. In this paper, a novel service cloud architecture is presented, and a linear regression model is used to predict the workload. Based on this predicted workload, an auto-scaling mechanism is proposed to scale virtual resources at different resource levels in service clouds. The auto-scaling mechanism combines the real-time scaling and the pre-scaling. Finally experimental results are provided to demonstrate that our approach can satisfy the user Service Level Agreement (SLA) while keeping scaling costs low.

JacUOD: A New Similarity Measurement for Collaborative Filtering

Collaborative filtering (CF) has been widely applied to recommender systems, since it can assist users to discover their favorite items. Similarity measurement that measures the similarity between two users or items is critical to CF. However, traditional similarity measurement approaches for memory-based CF can be strongly improved. In this paper, we propose a novel similarity measurement, named Jaccard Uniform Operator Distance (JacUOD), to effectively measure the similarity. Our JacUOD approach aims at unifying similarity comparison for vectors in different multidimensional vector spaces. Compared with traditional similarity measurement approaches, JacUOD properly handles dimension-number difference for different vector spaces. We conduct experiments based on the well-known MovieLens datasets, and take user-based CF as an example to show the effectiveness of our approach. The experimental results show that our JacUOD approach achieves better prediction accuracy than traditional similarity measurement approaches.

An adaptive prediction approach based on workload pattern discrimination in the cloud

Generally speaking, the workloads are changing rapidly on the Internet, but there is still regularity of changing patterns. Currently, workload prediction has become a promising tool to facilitate automatic scaling of resource management, and thus reducing the cost and improving resource utilization in the cloud. Most current predication methods of workload are based on a single model. However, because the network traffics are usually mixed and inseparable, it is hard to get the satisfactory prediction performance by means of a single model. To solve this problem, an adaptive approach for work load prediction is proposed in this paper. This approach firstly categorizes the workloads into different classes which are automatically assigned for different prediction models according to workload features. Furthermore, the workload classification problem is transformed into a task assignment one by establishing a mixed 01 integer programming model, and an online solution is provided. We used Google Cluster trace to evaluate the proposed approach. The experimental results demonstrate that the proposed approach improves the platform cumulative relative prediction errors by 29.06%, 8.42% and 40.86% respectively in comparison with the time-series prediction methods (Autoregressive Integrated Moving Average (ARIMA), Support Vector Machines (SVMs) and Linear Regression (LR).

Optimizing Workload Category for Adaptive Workload Prediction in Service Clouds

It is important to predict the total workload for facilitating auto scaling resource management in service cloud platforms. Currently, most prediction methods use a single prediction model to predict workloads. However, they cannot get satisfactory prediction performance due to varying workload patterns in service clouds. In this paper, we propose a novel prediction approach, which categorizes the workloads and assigns different prediction models according to the workload features. The key idea is that we convert workload classification into a 0–1 programming problem. We formulate an optimization problem to maximize prediction precision, and then present an optimization algorithm. We use real traces of typical online services to evaluate prediction method accuracy. The experimental results indicate that the optimizing workload category is effective and proposed prediction method outperforms single ones especially in terms of the platform cumulative absolute prediction error. Further, the uniformity of prediction error is also improved.

Personal Privacy Protection Framework Based on Hidden Technology for Smartphones

Smartphones are increasingly used for storing a large amount of sensitive used data. Users can protect their sensitive data through encryption and locking screen. A pattern lock is one of the ways to lock a screen. This method is used by most users because of its ease of use and memorization. However, a pattern lock with low security level is inadequate to protect the sensitive data of the user when it encounters a brute force or other physical attack (e.g., smudge attack). Furthermore, it bypasses all of the protection measures of mobile device when users are coerced into disclosing their passwords. Steganographic techniques and deniable encryption are designed to protect the sensitive data of the user as well as secure communications and can hide sensitive data on a disk or during communication with other devices. To overcome these deficiencies that mobile devices present, we present a novel, practical safe framework called MobiMimosa that is based on plausible deniable encryption. MobiMimosa enables multiple hidden encryption volumes and dynamic mounting of hidden volumes, which facilitates the transfer of sensitive data from a normal volume to a hidden volume. Simultaneously, to meet the personalized needs of the security of the mobile device, MobiMimosa enables a strategy to be set that can trigger the uninstalling of a sensitive app and the destruction of sensitive data. MobiMimosa also greatly alleviates corruption of the cross-volume boundary that is present in previous smartphone PDE schemes. We implemented a prototype system on the android device.

MobiGemini: sensitive-based data and resource protection framework for mobile device

With the popularity of smartphones and the rapid development of mobile internet, smartphone becomes an important tool that store sensitive data of owner. Encryption naturally becomes a necessary means of protection. In certain situations, this is inadequate, as user may be coerced to hand over decryption keys or passwords of sensitive APP (AliPay) on mobile device. Therefore, only encryption cannot protect sensitive APP and privacy data stored on user’s smartphone. To address these obstacles, we design a protection system called MobiGemini. It enables automatic uninstalling service that can immediately uninstall multiple APP at same time, and also enabling plausibly deniable encryption (PDE) on mobile devices by hiding encrypted volume within random data in free space of cache partition. We improve the key store way of previous PDE schemes on mobile device. The evaluation results show that the scheme introduces a few overhead compared with original android system enabling full disk encryption.

Design and implementation of SIP User-Agent based on applet

Multimedia conference system based on SIP is a real-time interactive communication service with multimedia applications, e.g., audio, video, whiteboard, and all users in the system should have the SIP clients which can create or receive SIP messages and send or display audio or video data for multimedia conference system. In this paper, the design and implementation of SIP User-Agent as applet for multimedia conference system is introduced which can run in Web browser. All modules of SIP UA are introduced separately, including user interface, call control module, media processing module and JAIN SIP stack. The test results prove that the system can correctly carry out multimedia communication.

Poster: Sdguard: An Android Application Implementing Privacy Protection and Ransomware Detection

Currently, the smartphone has become an essential communication and amusement tool, which has strong computing power and a variety of functions. Especially, the market share of smartphone with android system account for 84% in 2016[1]. Under android system, a large of privacy data (e.g. photos or videos) are stored in external storage (emulated Sdcard storage), which can be accessed by installed apps. This not only results in privacy leakage but also incurs ransomware attack[2] (e.g. simplocker). Therefore, we present Sdguard, an app, can implement fine-grain permission control based on Linux DAC mechanism and detect ransomware which encrypts content of file stored in external storage or lock user screen. To install Sdguard app, we need to ensure that the smartphone has been rooted and use FUSE filesystem on external storage. During installing, sdcard daemon of android (i.e. FUSE daemon) is replaced by our customized sdcard daemon. After rebooting system, the customized daemon is loaded, and each component of Sdguard is running.

Poster: EasyDefense: Towards Easy and Effective Protection Against Malware for Smartphones

As the dominant mobile operating system in the markets of smartphones, Android platform is increasingly targeted by attackers. Besides, attackers often produce novel malware to bypass the conventional detection approaches, which are largely reliant on expert analysis to design the discriminative features manually. Therefore, more effective and easy-to-use approaches for detection of Android malware are in demand. In this paper, we design and implement EasyDefense, a lightweight defense system that is integrated with Android OS for easy and effective detection of Android malware utilizing machine learning methods and the ensemble of them. Besides universal static features such as permissions and API calls, EasyDefense also employs the N-gram features of operation codes (opcodes). These N-gram features are extracted and learnt automatically from raw data of applications. Experimental results on 204,650 applications show that users can easily and effectively protect the privacy and security on their smartphones through this system.

Predicting of Job Failure in Compute Cloud Based on Online Extreme Learning Machine: A Comparative Study

Early prediction of job failures and specific disposal steps in advance could significantly improve the efficiency of resource utilization in large-scale data center. The existing machine learning-based prediction methods commonly adopt offline working pattern, which cannot be used for online prediction in practical operations, in which data arrive sequentially. To solve this problem, a new method based on online sequential extreme learning machine (OS-ELM) is proposed in this paper to predict online job termination status. With this method, real-time data are collected according to the sequence of job arriving, the job status could be predicted and the operation model is thus updated based on these data. The method with online incremental learning strategy has fast learning speed and good generalization. Comparative study using Google trace data shows that prediction accuracy of the proposed method is 93% with updating model in 0.01 s. Compared with some state-of-the-art methods, such, as support vector machine (SVM), ELM, and OS-SVM, the method developed in this paper has many advantages, such as less time-consuming in establishing and updating the model, higher prediction accuracy and precision, and better false negative performance.