Framework for Static Analysis of PHP Applications

Dynamic languages, such as PHP and JavaScript, are widespread and heavily used. They provide dynamic features such as dynamic type system, virtual and dynamic method calls, dynamic includes, and built-in dynamic data structures. This makes it hard to create static analyses, e.g., for automatic error discovery. Yet exploiting errors in such programs, especially in web applications, can have significant impacts. In this paper, we present static analysis framework for PHP, automatically resolving features common to dynamic languages and thus reducing the complexity of defining new static analyses. In particular, the framework enables defining value and heap analyses for dynamic languages independently and composing them automatically and soundly. We used the framework to implement static taint analysis for finding security vulnerabilities. The analysis has revealed previously unknown security problems in real application. Comparing to existing state-of-the-art analysis tools for PHP, it has found more real problems with a lower false-positive rate.