AXA: Cross-Language Analysis through Integration of Single-Language Analyses

Modern software is often implemented in multiple interacting programming languages. When performing static analysis of such software, it is desirable to reuse existing single-language analyses to allow access to the results of decades of implementation effort. However, there are major challenges for this approach. In this paper, we analyze them and present AXA, an architecture that addresses them and enables cross-language analysis by integrating single-language analyses. To evaluate AXA, we implemented a cross-language points-to analysis for Java applications that interact with native code via Java Native Interface (JNI) and with JavaScript code via Java's ScriptEngine. The evaluation shows that AXA enables significant reuse of existing static analyses. It also shows that AXA supports complex interactions and significantly increased recall of reused analyses without compromising precision.