Authentication in distributed systems: theory and practice

Reviewer: Stanley A. Kurzban

Although the use of distributed systems is growing rapidly, they have not yet come to dominate conventional business data processing. The authors usefully address a significant impediment to such domination: effective access control across distributed systems. Various open systems efforts have focused on functional requirements, and much work has been done on means of personal identification and on cryptographic protocols and algorithms that protect data used to authenticate identities. The authors provide a theoretical foundation for authentication in distributed systems. They give their audience of software designers and theoreticians the tools they need to develop precise statements about who and what represents whom in computerized transactions. Those notions are vital to implementation of effective, industrial-strength, identity-based access control, without which processing of organizations' most valuable data would be intolerably hazardous. The notion that a communication line may trustably represent a person is basic, but much more is involved. People may represent one another. Credentials must be fetched, presented, and authenticated. Names must be established, distinguished, and associated with one another. People must aggregate in groups and assume functional roles. The authors develop formalisms for all these tasks meticulously yet understandably. Regrettably, the authors do little to relate their work to the world of business data processing. They write only of system administrators, when the world also needs to know about clerks and managers. The only type of access that appears in their examples is read access, but write access and especially distinctions between the two are far more important. They barely begin to relate their work to that of Clark and Wilson [1], Chaum [2], and Abrams [3], all of which it can serve well. They hardly touch on such vital closely related notions as object naming, auditing, identity-verification devices, and capabilities, although they do refer the reader to longer works they have written that might extend to some of these topics. Criticism about what the authors have not written must not obscure what they have, however. This paper is a significant contribution to the field, and those interested in the subject will find it an invaluable foundation for their work.