Separating two roles of hashing in one-way message authentication

We analyse two new and related families of one-way authentication protocols, where a party wants to authenticate its public information to another. In the first, the objective is to do without shared passwords or a PKI, making use of low-bandwidth empirical/authentic channels where messages cannot be faked or modified. The analysis of these leads to a new security principle, termed separation of security concerns, under which protocols should be designed to tackle one-shot attacks and combinatorial search separately. This also leads us develop a new class of protocols for the case such as PKI where a relatively expensive signature mechanism exists. We demonstrate as part of this work that a popular protocol in the area, termed MANA I, neither optimises human effort nor offers as much security as had previously been believed. We offer a number of improved versions for MANA I that provides more security for half the empirical work, using a more general empirical channel.