Experience report: an empirical study of PHP security mechanism usage

Abstract

The World Wide Web mainly consists of web applications written in weakly typed scripting languages, with PHP being the most popular language in practice. Empirical evidence based on the analysis of vulnerabilities suggests that security is often added as an ad-hoc solution, rather than planning a web application with security in mind during the design phase. Although some best-practice guidelines emerged, no comprehensive security standards are available for developers. Thus, developers often apply their own favorite security mechanisms for data sanitization or validation to prohibit malicious input to a web application. In the context of our development of a new static code analysis tool for vulnerability detection, we studied commonly used input sanitization or validation mechanisms in 25 popular PHP applications. Our analysis of 2.5 million lines of code and over 26 thousand secured data flows provides a comprehensive overview of how developers utilize security mechanisms in practice regarding different markup contexts. In this paper, we discuss these security mechanisms in detail and reveal common pitfalls. For example, we found certain markup contexts and security mechanisms more frequently vulnerable than others. Our empirical study helps researchers, web developers, and tool developers to focus on error-prone markup contexts and security mechanisms in order to detect and mitigate vulnerabilities.