Generalizing Access Control in Class-Based Languages

Class-based languages, such as Java and C#, implement encapsulation via access modifiers on classes, fields and methods. These modifiers can be viewed as a form of access control, where the principals are units of code. Each language supports a fixed set of these modifiers (e.g. public, class-private, package-private, protected in Java) making the access control policy fixed. For more dynamic controls, these languages rely on libraries (e.g. JAAS in Java) which provide other notions of principal such as "codebase location", "author " and "user". The libraries are substantially more difficult to use than the language-level mechanisms, although they serve at least partially the same function. In this paper we generalize the access control model for class-based languages to one that is more flexible and that can be expressed formally. We then show how this more-expressive model removes some limitations from the programmer. Our extended language is implemented in an interpreter which implements the operational semantics showing access control errors at runtime..