Mathematical Models of Mobile Computing Devices and Application Software

点击翻译按钮获取中文摘要

The widespread use of mobile devices and applications (apps) is translating into large-scale economic benefits for Canadians. According to a 2016 report prepared by the market research firm Nordicity for the Canadian Wireless Telecommunications Association, companies in the Canadian wireless communications ecosystem generated $48.96 billion in revenue in 2015. Smartphones, embedded devices, wearable devices, and wireless sensors are examples of mobile devices. They play key roles in emerging applications: personal communication, financial transactions, asset monitoring, autonomous vehicles, drone-based monitoring of national borders and conflict zones, and sensing and communications performed by the armed forces. Mobile devices have become increasingly powerful, with feature-rich operating systems, multi-core processors, gigabytes of memory, multi-radio interfaces, and an array of sensors. However, unlike desktops, mobile devices run on batteries and operate in harsh communication environments. Among the key challenges in designing robust mobile apps are modeling and identification of malicious code, modeling of the power cost of individual apps and devices, designing test suites for performance testing, and extending battery life. Through the proposed research program, we will pursue the following objectives: (i) determine if mobile apps are engaging in suspicious activities; (ii) assuming that a device is running a known set of apps, externally determine if it is running some unknown apps; (iii) estimate the loss of energy due to suspicious activities; (iv) given the operational model of an app and a desired level of user's quality of experience (QoE), compute the app's configuration parameters (ACPs) and the network's configuration parameters (NCPs) so that when the app is configured with the selected ACPs and executed under the network's selected NCPs, it delivers the desired QoE; the set of ACPs, NCPs, and QoE for a given app constitute a single test case for performance testing; and (v) develop test coverage criteria to adequately cover the app, ACPs, NCPs, and QoE. Our methodologies to achieve the objectives include: external measurement of power, application of independent component analysis and data disaggregation techniques in the identification of malicious apps; modeling of transport protocols over wireless interfaces; designing performance models of apps by combining the protocol models and Markov models of apps; generating test cases from performance models by developing a concept of model inversion; and designing selection criteria for performance test adequacy. The models, algorithms, and tools to be developed in this research will find applications in the design and test of robust mobile devices for security, safety, and business critical applications developed in the Canadian mobile device ecosystem for both commercial and defense applications.