Date

2014

Document Type

Dissertation

Degree

Doctor of Philosophy

Department

Computer Science

First Adviser

Tan, Gang

Other advisers/committee members

Chuah, Mooi-Choo; Davison, Brian D.; Liu, Yu David

Abstract

A Foreign Function Interface (FFI) is a mechanism that allows software written in one host programming language to directly use another foreign programming language by invoking function calls across language boundaries. Today's software development often utilizes FFIs to reuse software components. Examples of such systems are the Java Development Kit (JDK), Android mobile OS, and Python packages in the Fedora LINUX operating systems. The use of FFIs, however, requires extreme care and can introduce undesired side effects that degrade software quality. In this thesis, we aim to improve several quality aspects of software composed of FFIs by applying static analysis. The thesis investigates several particular characteristics of FFIs and studies software bugs caused by the misuse of FFIs. We choose two FFIs, the Java Native Interface (JNI) and the Python/C interface, as the main subjects of this dissertation. To reduce software security vulnerabilities introduced by the JNI, we first propose definitions of new patterns of bugs caused by the improper exception handlings between Java and C. We then present the design and implement a bug finding system to uncover these bugs. To ensure software safety and reliability in multithreaded environment, we present a novel and efficient system that ensures atomicity in the JNI. Finally, to improve software performance and reliability, we design and develop a framework for finding errors in memory management in programs written with the Python/C interface. The framework is built by applying affine abstraction and affine analysis of reference-counts of Python objects. This dissertation offers a comprehensive study of FFIs and software composed of FFIs. The research findings make several contributions to the studies of static analysis and to the improvement of software quality.

Share

COinS