SHF: extreme semantic dynamic analysis of multilingual code

The National Science Foundation Division of Information and Intelligent Systems awarded The Research Foundation For The State University of New York $615,000 on January 1, 2027, under the Computer and Information Science and Engineering program (CFDA 47.070) to develop extreme semantic dynamic analysis techniques for multilingual software systems.

The project, running through December 31, 2029, with performance in Buffalo, New York, focuses on revealing causal relationships between computations across programming language boundaries. The research introduces semantic dependence analysis—identifying when changing one program point causally affects later computations during execution—which differs from traditional structural dependency models that capture connections without necessarily explaining causality. The work employs extreme value perturbations, including nullification of selected computations, and compares original and perturbed executions to identify causal effects. The project develops semantic information-flow analysis to trace how information propagates from sources to sinks across multilingual software components, implements incremental algorithms that update analysis results as code evolves, and produces tools, benchmarks, and educational materials to help developers build more reliable and secure software. Evaluation occurs using controlled benchmarks, real multilingual systems, and vulnerability-detection tasks.

The Research Foundation serves as fiscal agent and sponsored programs administrator for the State University of New York system; the award flows through its parent entity, The Research Foundation For The State University of New York (Sponsored Programs Office - System Administration).

Abstract

Software is often assembled from parts written in different programming languages. This design helps developers combine ease of use, speed, and reuse, but IT also makes failures harder to explain. A problem may appear in one language even though its real cause lies in another. Existing tools can often show that pieces of code are connected, but a connection is not the same as cause. This project studies how to reveal which earlier computations cause later values, behaviors, or security-relevant effects in running multilingual software. The project's novelties are controlled perturbations of selected computations and new methods for observing how their effects ripple across language boundaries. The project's broader significance and importance are better explanations for software failures and vulnerabilities, leading to tools, benchmarks, and educational materials that help developers build more reliable and secure software. The project develops extreme semantic dynamic analysis for multilingual code. The central idea is semantic dependence: one program point depends on another when changing the earlier computation changes the later computation during execution. This differs from traditional data and control dependence, which captures structural relationships but not necessarily causality. The project introduces extreme value perturbations, including nullification of selected computations, and compares original and perturbed executions to identify causal effects. These dependencies support semantic information-flow analysis, which explains how information propagates from sources to sinks across components written in different languages. The project also develops incremental algorithms that update prior analysis results as software evolves, reducing repeated analysis cost. The research is evaluated using controlled benchmarks, real multilingual systems, and vulnerability-detection tasks. The outcomes include new models, algorithms, open-source tools, and benchmark suites for causal analysis of modern multilingual software. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.