Csiszar and Narayan[3] defined the notion of secret key capacity for multiple terminals, characterized it as a linear program with Slepian-Wolf constraints of the related source coding problem of communication for omniscience, and upper bounded it by some information divergence expression from the joint to the product distribution of the private observations. This paper proves that the bound is tight for the important case when all users are active, using the polymatroidal structure[6] underlying the source coding problem. When some users are not active, the bound may not be tight. This paper gives a counter-example in which 3 out of the 6 terminals are active.
Csiszar and Narayan[3] defined the notion of secret key capacity for multiple terminals, characterized it as a linear program with Slepian-Wolf constraints of the related source coding problem of communication for omniscience, and upper bounded it by some information divergence expression from the joint to the product distribution of the private observations. This paper proves that the bound is tight for the important case when all users are active, using the polymatroidal structure[6] underlying the source coding problem. When some users are not active, the bound may not be tight. This paper gives a counter-example in which 3 out of the 6 terminals are active.
We study the problem of information-theoretically secure secret key agreement under the well-known source model and channel model. In both of these models, multiple terminals wish to create a shared secret key that is secure from a passive eavesdropper. The terminals have access to a noiseless public communication channel and an additional resource that depends on the model. In the source model, the resource is an external source that repeatedly beams correlated randomness to the terminals; whereas in the channel model, the resource is a secure but noisy discrete memoryless broadcast channel. We derive new lower and upper bounds on the secret key capacity under both the source model and the channel model. The technique used for deriving our bound for the source model is to find certain properties of functions of joint probability distributions which, applied to the joint distribution of the source, will imply that they dominate the secret key capacity, and then prove the bound by a verification argument. A similar technique is used for the channel model. Finally, we also define a problem of communication for omniscience by a neutral observer and establish the equivalence between this new problem and the problem of secret key agreement. This generalizes an earlier result of Csiszár and Narayan.
This dissertation is primarily concerned with the communication required to achieve secret key (SK) capacity in a multiterminal source model. The multiterminal source model introduced by Csiszar and Narayan consists of a group of remotely located terminals with access to correlated sources and a noiseless public channel. The terminals wish to secure their communication by agreeing upon a group secret key. The key agreement protocol involves communicating over the public channel, and agreeing upon an SK secured from eavesdroppers listening to the public communication. The SK capacity, i.e., the maximum rate of an SK that can be agreed upon by the terminals, has been characterized by Csiszar and Narayan. Their capacity-achieving key generation protocol involved terminals communicating to attain omniscience, i.e., every terminal gets to recover the sources of the other terminals. While this is a very general protocol, it often requires larger rates of public communication than is necessary to achieve SK capacity.
The primary focus of this dissertation is to characterize the communication complexity, i.e., the minimum rate of public discussion needed to achieve SK capacity. A lower bound to communication complexity is derived for a general multiterminal source, although it turns out to be loose in general. While the minimum rate of communication for omniscience is always an upper bound to the communication complexity, we derive tighter upper bounds to communication complexity for a special class of multiterminal sources, namely, the hypergraphical sources. This upper bound yield a complete characterization of hypergraphical sources where communication for omniscience is a rate-optimal protocol for SK generation, i.e., the communication complexity equals the minimum rate of communication for omniscience.
Another aspect of the public communication touched upon by this dissertation is the necessity of omnivocality, i.e., all terminals communicating, to achieve the SK capacity. It is well known that in two-terminal sources, only one terminal communicating success to generate a maximum rate secret key. However, we are able to show that for three or more terminals, omnivocality is indeed required to achieve SK capacity if a certain condition is met. For the specific case of three terminals, we show that this condition is also necessary to ensure omnivocality is essential in generating a SK of maximal rate. However, this condition is no longer necessary when there are four or more terminals.
A certain notion of common information, namely, the Wyner common information, plays a central role in the communication complexity problem. This dissertation thus includes a study of multiparty versions of the two widely used notions of common information, namely, Wyner common information and Gacs-Korner (GK) common information. While evaluating these quantities is difficult in general, we are able to derive explicit expressions for both types of common information in the case of hypergraphical sources.
We also study fault-tolerant SK…
A mutual dependence expression is established for the secret key agreement problem when all users are active. In certain source networks, the expression can be interpreted as certain notions of connectivity and network information flow. In particular, the secrecy problem can be mapped to a new class of network coding problems with selectable links and undirected broadcast links. For such networks, the secrecy capacities serve as upper bounds on the maximum network throughputs, while the network coding solutions can be used for secret key agreement.
We derive single-letter characterizations of (strong) secrecy capacities for models with an arbitrary number of terminals, each of which observes a distinct component of a discrete memoryless multiple source, with unrestricted and interactive public communication permitted between the terminals. A subset of these terminals can serve as helpers for the remaining terminals in generating secrecy. According to the extent of an eavesdropper's knowledge, three kinds of secrecy capacity are considered: secret key (SK), private key (PK), and wiretap secret key (WSK) capacity. The characterizations of the SK and PK capacities highlight the innate connections between secrecy generation and multiterminal source coding without secrecy requirements. A general upper bound for WSK capacity is derived which is tight in the case when the eavesdropper can wiretap noisy versions of the components of the underlying multiple source, provided randomization is permitted at the terminals. These secrecy capacities are seen to be achievable with noninteractive communication between the terminals. The achievability results are also shown to be universal.
Shannon-theoretic secret key generation by several parties is considered for source models in which the distinct components of a multiple source observed separately by multiple terminals, and for channel models in which a secure noisy channel with one input terminal and multiple output terminals, and, additionally in both cases, a public noiseless channel of unlimited capacity, are available for accomplishing this goal. The secret key is generated for a set A of terminals, with the remaining terminals (if any) cooperating in this task through their public communication. We show that for source models in which secrecy is required from an eavesdropper that observes only the public communication and perhaps also a set of terminals disjoint from A, secrecy capacity can be achieved with noninteractive communication, the key being generated by any chosen terminal in the secret key-seeking set A of terminals obliviously of the public communication. For models in which the eavesdropper also possesses side information that is not available to any of the terminals cooperating in secrecy generation, an upper bound for the secrecy capacity and a sufficient condition for its tightness are given. The latter partially fills a gap in the authors' previous work [6].