Ten years ago, Google released the first version of its new operating system: Android. With an open market for third party applications, attackers started to develop malicious applications. Researchers started new works too. Inspired by previous techniques for Windows or GNU/Linux malware, a lot of papers introduced new ways of detecting, classifying, defeating Android malware. In this paper, we propose to explore the technical difficulties of experimenting with Android malware. These difficulties are encountered by researchers, each time they want to publish a solid experiment validating their approach. How to choose malware samples? How to process a large amount of malware? What happens if the experiment needs to execute dynamically a sample? The end of the paper presents the upcoming scientific challenges of the community interested in malware analysis.