Android'de Çökme Tespitini Iyileştirme Amaçli Model-tabanli ve Rastgele Karma Yöntem

Android applications are widely used around the world. Most of these applications contain potential crashes. Many recent academic studies focus on black-box testing of Android applications to detect these crashes. A simple random testing tool, Monkey, detects more crashes than the state-of-the-art black-box testing tools, but can not reach some activities that are located deep within the application. We propose an hybrid approach that combines our model-learning tool, AndroFrame, and Monkey. With this hybrid approach, we aim to increase activity coverage and improve crash detection. We conduct experiment on 20 Android applications. As a result, our hybrid approach achieves 2% more activity coverage and detects 21 more crashes compared to AndroFrame. Compared to Monkey, our hybrid approach achieves 24% more coverage and detects 5 more crashes.