Version Detection for Software Discovery in the Cloud

Abstract

With the growth in server traffic and component diversity in cloud systems, administrators face the increasingly onerous task of monitoring system activity. Failure to keep track of the contents of virtual servers can limit overall efficiency and create security risks for users. Prior work in software discovery attempted to address this problem by identifying applications based on file system activity. While some of these methods have claimed to be extensible to detection of specific versions of an application, version detection has yet to be demonstrated. In this paper, we propose version detection algorithms that operate on top of Praxi, an existing open-source software discovery tool. These algorithms introduce a rule-based component to differentiate between versions, whose file system footprints can appear very similar. We find that our best method achieves up to 99.9% accuracy in version detection experiments compared to Praxi's original 94% accuracy, albeit at the cost of increased runtime. This work confirms the feasibility of version detection in software discovery and provides a starting point for implementing this feature in software discovery tools.