Jean Yang

Assistant Professor (on leave)
Computer Science Department
Carnegie Mellon University

Administrative Assistant

Erin Davis, eedavis [at] andrew [dot] cmu [dot] edu

Research

Active Research Directions

Our group is currently investigating how the following two programming models can change the way people think about their respective domains, and facilitate the creation of programs and analyses that were previously difficult or impossible.

• Policy-agnostic programming for security and privacy. As an alternative to approaches for detecting information leaks, my collaborators and I propose a new programming model that factors out the specification of security and privacy concerns from the rest of the program and enforces the properties by construction. In our work on Jeeves we designed a language semantics for policy-agnostic programming with informaton flow policies and developed dynamic and static enforcement techniques. We are currently interested in (1) extending the approach for statistical privacy and (2) techniques for retrofitting legacy code with the policy-agnostic model, for purposes of fixing bugs and interacting with new policies and code.
• Rule-based programming for biological modeling. Traditionally, researchers model intracellular signalling using systems of ordinary differential equations (ODEs), but there are two problems with ODE models. First, a precise model requires a different ODE for each interaction between agents, causing ODE models to scale poorly with respect to number of agent types. Second, ODE models have little structure that we can exploit for scale-mitigating analyses. As an alternative, rule-based languages allow the representation of models as programs describing rewrites over graphs, where nodes correspond to proteins and edges describe protein complexes. Not only are these programs more concise than the corresponding ODE systems, but their structure also supports various analyses that are otherwise not possible. Our current work focuses on analyzing causal relationships between rules, and combining causal information with language design and model-checking techniques to create biologically relevant model analyses.

Publications

Workshop Talks

Software Patents