A response to “can we eliminate certificate revocation lists?”

Abstract

The massive growth of electronic commerce on the Internet heightens concerns over the lack of meaningful certificate management. One issue limiting the availability of such services is the absence of scalable certificate revocation. The use of certificate revocation lists (CRLs) to convey revocation state in public key infrastructures has long been the subject of debate. Centrally, opponents of the technology attribute a range of semantic and technical limitations to CRLs. In this paper, we consider arguments advising against the use of CRLs made principally by Rivest in his paper “Can we eliminate certificate revocation lists?” [1]. Specifically, the assumptions and environments on which these arguments are based are separated from those features inherent to CRLs. We analyze the requirements and potential solutions for three distinct PKI environments. The fundamental tradeoffs between revocation technologies are identified. Prom the case study analysis we show how, in some environments, CRLs are the most efficient vehicle for distributing revocation state. The lessons learned from our case studies are applied to a realistic PKI environment. The result, revocation on demand, is a CRL based mechanism providing timely revocation information.