Establishing the broadcast efficiency of the Subset Difference Revocation Scheme

Abstract

When an organisation chooses a system to make regular broadcasts to a changing user base, there is an inevitable trade off between the number of keys a user must store and the number of keys used in the broadcast. The Complete Subtree and Subset Difference Revocation Schemes were proposed as efficient solutions to this problem. However, all measurements of the broadcast size have been in terms of upper bounds on the worst-case. Also, the bound on the latter scheme is only relevant for small numbers of revoked users, despite the fact that both schemes allow any number of such users. Since the broadcast size can be critical for limited memory devices, we aid comparative analysis of these important techniques by establishing the worst-case broadcast size for both revocation schemes.