The communication complexity of achieving secret key (SK) capacity in the multiterminal source model of Csiszár and Narayan is the minimum rate of public communication required to generate a maximal-rate SK. It is well known that the minimum rate of communication for omniscience, denoted by R CO , is an upper bound on the communication complexity, denoted by R SK . A source model for which this upper bound is tight is called R SK -maximal. In this paper, we establish a sufficient condition for R SK -maximality within the class of pairwise independent network (PIN) models defined on hypergraphs. This allows us to compute R SK exactly within the class of PIN models satisfying this condition. On the other hand, we also provide a counterexample that shows that our condition does not in general guarantee R SK -maximality for sources beyond PIN models.
I. INTRODUCTIONCsiszár and Narayan [6] introduced the problem of secret key (SK) generation within the multiterminal i.i.d. source model. In this model, there are multiple terminals, each of which observes a distinct component of a source of correlated randomness. The goal is for the terminals to agree on a shared SK via communication over an insecure noiseless public channel. The SK is to be secured from passive eavesdroppers with access to the public channel. The maximum rate of such an SK, i.e. the SK capacity, was characterized in [6], and a protocol for attaining SK capacity was given, which involved communication for omniscience, i.e., all terminals recovering the entire information of all the other terminals. However, it was pointed out (see remark following Theorem 1 in [6]) that omniscience is not always necessary for achieving SK capacity. A question that naturally arises is the following (see [6, Section VI] and [12, Section V]): what is the minimum rate of public communication required to achieve SK capacity? We call this minimum rate of public communication the communication complexity 1 of achieving SK capacity, and denote it by R SK . The protocol from [6] shows that R SK is upper bounded by the minimum rate of public communication required for omniscience, denoted by R CO . We refer to sources for which this upper bound is tight as R SK -maximal.There have been a few attempts at characterizing R SK . In [13, Theorem 3] Tyagi has completely characterized the † M. Mukherjee and N. Kashyap are with the 1 Our use of "communication complexity" differs from the use prevalent in the theoretical computer science literature where, following [15], it refers to the total amount of communication, in bits, required to perform some distributed computation. communication complexity for two terminals in terms of an interactive common information, a type of Wyner common information [14]. Our previous work [10] involved extension of Tyagi's results to the case of m > 2 terminals. Specifically, we gave a lower bound [10, Theorem 2] on the communication complexity using a multiterminal variant of Tyagi's interactive common information. We were able to evaluate this lower bou...