Randomized Event Sequence Generation Strategies for Automated Testing of Android Apps

Abstract

Mobile apps are often tested with automatically generated sequences of Graphical User Interface (GUI) events. Dynamic GUI testing algorithms construct event sequences by selecting and executing events from GUI states at runtime. The event selection strategy used in a dynamic GUI testing algorithm may directly influence the quality of the test suites it produces. Existing algorithms use a uniform probability distribution to randomly select events from each GUI state and they are often not directly applicable to mobile apps. In this paper, we develop a randomized algorithm to dynamically construct test suites with event sequences for Android apps. We develop two frequency-based event selection strategies as alternatives to uniform random event selection. Our event selection algorithms construct event sequences by dynamically altering event selection probabilities based on the prior selection frequency of events in each GUI state. We compare the frequency-based strategies to uniform random selection across nine Android apps. The results of our experiments show that the frequency-based event selection strategies tend to produce test suites that achieve better code coverage and fault detection than test suites constructed with uniform random event selection.