Monkey Gamer: Automatic profiling of Android games

Creation of smartphone applications has undergone a massive explosion in recent years and there is an urgent need for evaluation of their resource efficiency, trustworthiness and reliability. A large proportion of these apps are going to be within the gaming area. In this paper we classify game apps on the basis of their development process, their I/O process and their interaction level. We present Monkey Gamer, a software to automatically play a large class of Android games and collect execution traces, based on a state machine to partially describe the game structure and interactions. A significant similarity is shown when comparing the results obtained by the Monkey Gamer and by human players, for three of the most popular Android games. We evaluate the performance of the Monkey Gamer by comparing the traces it generates with traces created when humans play the games, and find significant similarity in the trace sets.