Characterizing Co-located Insecure Coding Patterns in Infrastructure as Code Scripts

Abstract

Context: Insecure coding patterns (ICPs), such as hard-coded passwords can be inadvertently introduced in infrastructure as code (IaC) scripts, providing malicious users the opportunity to attack provisioned computing infrastructure. As performing code reviews is resource-intensive, a characterization of co-located ICPs, i.e., ICPs that occur together in a script can help practitioners to prioritize their review efforts and mitigate ICPs in IaC scripts. Objective: The goal of this paper is to help practitioners in prioritizing code review efforts for infrastructure as code (IaC) scripts by conducting an empirical study of co-located insecure coding patterns in IaC scripts. Methodology: We conduct an empirical study with 1613, 2764 and 2845 Puppet scripts respectively collected from three organizations namely, Mozilla, Openstack, and Wikimedia. We apply association rule mining to identify co-located ICPs in IaC scripts. Results: We observe 17.9%, 32.9%, and 26.7% of the scripts to include colocated ICPs respectively, for Mozilla, Openstack, and Wikimedia. The most frequent co-located ICP category is hard-coded secret and suspicious comment. Conclusion: Practitioners can prioritize code review efforts for IaC scripts by reviewing scripts that include co-located ICPs.