Abstract
With the advances of wireless communication and the popularity of smart phones, more and more people access the Internet and the WWW by their mobile devices. For accessing the WWW, browser is the primary interface, and thus the performance and behavior of the browser become critical for the design of mobile devices. In this thesis, we are interested in the internal behavior of the browser for processing web pages and applications. Existing profiling tools only provide limited information such as memory and CPU cycles consumed. In this thesis, we develop a profiling tool that allows us to understand internal behavior of the browser by collecting information from different layers of the system: virtual machine, operating system, and hardware. To characterize the behavior of browsers, independent of their implementation, we propose a state-machine model that describes the workflow of a generic browser in processing a web page. We then classify and summarize the data collected from our tool according to the state-machine model to reveal its internal behavior. We evaluate our method and tool using real web pages and show how that the browser can be classified into different stages and how the collected data are be mapped to these stages to characterize the browser behavior.