CrawlDroid

Abstract

This paper presents an effective model-based GUI testing technique for Android apps. To avoid local and repetitive exploration, our approach groups equivalent widgets in a state and designs a novel feedback-based exploration strategy, which dynamically adjusts the priority of actions based on the execution result of those already triggered ones, and tends to select actions that can reach news states of apps. We implemented our technique in a tool, called CrawlDroid, and conducted empirical experiments. Our results show that the proposed technique is effective, and covers more code within a fixed testing budget.