A type-based approach to program security

Abstract

This paper presents a type system which guarantees that well-typed programs in a procedural programming language satisfy a noninterference security property. With all program inputs and outputs classified at various security levels, the property basically states that a program output, classified at some level, can never change as a result of modifying only inputs classified at higher levels. Intuitively, this means the program does not "leak" sensitive data. The property is similar to a notion introduced years ago by Goguen and Meseguer to model security in multi-level computer systems. We also give an algorithm for inferring and simplifying principal types, which document the security requirements of programs.