Efficient Type Inference for Secure Information Flow
Katia Hristova, Tom Rothamel, Yanhong A. Liu, and Scott D. Stoller.

This paper describes the design, analysis, and implementation of an efficient algorithm for information flow analysis expressed using a type system. Given a program and an environment of security classes for information accessed by the program, the algorithm checks whether the program is well typed, i.e., there is no information of higher security classes flowing into places of lower security classes, by inferring the highest or lowest security class as appropriate for each program node. We express the analysis as a set of Datalog-like rules based on the typing and subtyping rules, and we use a systematic method to generate specialized algorithms and data structures directly from the Datalog-like rules. The generated implementation traverses the program in two passes and uses a combination of linked and indexed data structures to represent program nodes, environments, and types. The time complexity of the algorithm is linear in the size of the input program, plus a small overhead for preprocessing the security classes. This complexity is confirmed through our prototype implementation and experimental evaluation on code generated from high-level specifications for real systems.


Scott Stoller's Home Page