An Evolutionary Approach for System Testing of Android Applications.

摘要

Mobile app markets have created a fundamental shift in the way software is delivered to the consumers. The benefits of this software supply model are plenty, including the ability to rapidly and effectively deploy, maintain, and enhance software used by the consumers. This paradigm, however, has given rise to a new set of concerns. Small organizations do not have the resources to sufficiently test their products, thereby defective apps are made available to the consumers of these markets. The situation is likely to exacerbate given that mobile apps are poised to become more complex and ubiquitous.;Automated testing of Android apps is impeded by the fact that they are built using an application development framework (ADF). ADF allows the programmers to extend the base functionality of the platform using a well-defined API. ADF also provides a container to manage the lifecycle of components comprising an app and facilitates the communication among them. As a result, unlike a traditional monolithic software system, an Android app consists of code snippets that engage one another using the ADF's sophisticated event delivery facilities. This hinders automated testing, as the app's control flow frequently interleaves with the ADF. At the same time, reliance on a common ADF provides a level of consistency in the implementation logic of apps that can be exploited for automating the test activities, as illustrated in this research.;Evolutionary testing technique has shown to be particularly effective for event driven software. In this research, I present the first evolutionary testing framework targeted at Android, called EvoDroid. Evolutionary testing is a form of search-based testing, where an individual corresponds to a test case, and a population comprised of many individuals is evolved according to certain heuristics to maximize the code coverage.;The most notable contribution of EvoDroid is its ability to overcome the common shortcoming of using evolutionary techniques for system testing. Evolutionary testing techniques are typically limited to local or unit level testing, as for system testing, they generate many invalid test cases. This occurs when the individuals (test cases) are crossed over to create new ones. The crossover does not consider which input and event genes are coupled to which part of the app, hence, it is not able to preserve the genetic makeup of parents in any meaningful way. I overcame this challenge by leveraging the knowledge of how Android specifies and constrains the way apps can be built. I devised a technique to analyze apps and infer models of their interface and behavior. Using these models my technique generates test cases reaching deep into the code in segments, i.e., sections of code that are amenable to evolutionary testing without the possibility of generating invalid test cases. Since a key concern in search-based testing is the execution time of the algorithm, I built EvoDroid to run the test cases in parallel on Android emulators deployed on the cloud, thus achieving several orders of magnitude improvement in execution time.