Taint analysis has many important applications such as software attacks prevention, data lifetime analysis, and information flow control. However, static taint analysis, in many scenarios, is not precise enough for certain applications such as attack provenance analysis, reverse engineering, and malware analysis. On the other hand, the high runtime overhead imposed by dynamic taint analysis makes its deployment impractical in production systems.

In this talk, I will present a hybrid static and dynamic taint analysis technique which improves the performance of dynamic taint analysis and yet is more precise than static taint analysis. We perform very lightweight logging of program execution flow to reconstruct the executed code, on which we perform pipelined segmented symbolic taint analysis. Our preliminary experiments show that our prototype implementation can achieve better performance with comparable precision.

Dinghao Wu is an Assistant Professor in the College of Information Sciences and Technology at The Pennsylvania State University. He received his Ph.D. in Computer Science from Princeton University in 2005. He was a research engineer at Microsoft in the Center for Software Excellence and the Windows Azure Division before joined Penn State. Dinghao does research on software systems, including software security, analysis, verification, software engineering, and programming languages. He has worked on foundational proof-carrying code, typed assembly languages, program analysis, and software and systems security projects. His current projects include lock-free concurrent security monitoring, real-time concurrent information flow tracking, and semantics-based software plagiarism detection. He also leads a project on cloud computing for energy and environmental sustainability.