International Association
for Cryptologic Research

In this paper we revisit the notion of generalized universal composability (GUC)

introduced by Canetti, Dodis, Pass and Walfish in 2007. The GUC model was

intended to model a practical setting where setup parameters, like a PKI or a CRS,

are made public once and for all and then used by many different protocols.

We show that there exist protocols which can be proven secure in the GUC model,

but which are obviously insecure in practice, in the setting that the GUC model was

intended to capture. We then proceed to revise the GUC model to a version that

better models the intended practical setting. We call the new notion strong generalized

UC. We finally prove that the GUC protocols proposed by Canetti, Dodis, Pass and Walfish

are also strong GUC secure, i.e., whereas there is a problem with the model, the

protocols seem to be secure in the intended setting.