Practical implementations of secret-key generation are often based on sequential strategies, which handle reliability and secrecy in two successive steps, called reconciliation and privacy amplification. In this paper, we propose an alternative approach based on polar codes that jointly deals with reliability and secrecy. Specifically, we propose secret-key capacity-achieving polar coding schemes for the following models: (i) the degraded binary memoryless source (DBMS) model with rate-unlimited public communication, (ii) the DBMS model with one-way rate-limited public communication, (iii) the 1-to-m broadcast model and (iv) the Markov tree model with uniform marginals. For models (i) and (ii) our coding schemes remain valid for non-degraded sources, although they may not achieve the secret-key capacity. For models (i), (ii) and (iii), our schemes rely on pre-shared secret seed of negligible rate; however, we provide special cases of these models for which no seed is required. Finally, we show an application of our results to secrecy and privacy for biometric systems. We thus provide the first examples of low-complexity secret-key capacity-achieving schemes that are able to handle vector quantization for model (ii), or multiterminal communication for models (iii) and (iv).
Practical implementations of secret-key generation are often based on sequential strategies, which handle reliability and secrecy in two successive steps, called reconciliation and privacy amplification. In this paper, we propose an alternative approach based on polar codes that jointly deals with reliability and secrecy. Specifically, we propose secret-key capacity-achieving polar coding schemes for the following models: (i) the degraded binary memoryless source (DBMS) model with rate-unlimited public communication, (ii) the DBMS model with one-way rate-limited public communication, (iii) the 1-to-m broadcast model and (iv) the Markov tree model with uniform marginals. For models (i) and (ii) our coding schemes remain valid for non-degraded sources, although they may not achieve the secret-key capacity. For models (i), (ii) and (iii), our schemes rely on pre-shared secret seed of negligible rate; however, we provide special cases of these models for which no seed is required. Finally, we show an application of our results to secrecy and privacy for biometric systems. We thus provide the first examples of low-complexity secret-key capacity-achieving schemes that are able to handle vector quantization for model (ii), or multiterminal communication for models (iii) and (iv).
Wyner's work on wiretap channels and the recent works on information theoretic security are based on random codes. Achieving information theoretical security with practical coding schemes is of definite interest. In this note, the attempt is to overcome this elusive task by employing the polar coding technique of Arıkan. It is shown that polar codes achieve non-trivial perfect secrecy rates for binary-input degraded wiretap channels while enjoying their low encoding-decoding complexity. In the special case of symmetric main and eavesdropper channels, this coding technique achieves the secrecy capacity. Next, fading erasure wiretap channels are considered and a secret key agreement scheme is proposed, which requires only the statistical knowledge of the eavesdropper channel state information (CSI). The enabling factor is the creation of advantage over Eve, by blindly using the proposed scheme over each fading block, which is then exploited with privacy amplification techniques to generate secret keys.
This dissertation deals with a multiterminal source model for secret key generation by multiple network terminals with prior and privileged access to a set of correlated signals complemented by public discussion among themselves. Emphasis is placed on a characterization of secret key capacity, i.e., the largest rate of an achievable secret key, and on algorithms for key construction. Various information theoretic security requirements of increasing stringency: weak, strong and perfect secrecy, as well as different types of sources: finite-valued and continuous, are studied. Specifically, three different models are investigated. First, we consider strong secrecy generation for a discrete multiterminal source model. We discover a connection between secret key capacity and a new source coding concept of "minimum information rate for signal dissemination," that is of independent interest in multiterminal data compression. Our main contribution is to show for this discrete model that structured linear codes suffice to generate a strong secret key of the best rate. Second, strong secrecy generation is considered for models with continuous observations, in particular jointly Gaussian signals. In the absence of suitable analogs of source coding notions for the previous discrete model, new techniques are required for a characterization of secret key capacity as well as for the design of algorithms for secret key generation. Our proof of the secret key capacity result, in particular the converse proof, as well as our capacity-achieving algorithms for secret key construction based on structured codes and quantization for a model with two terminals, constitute the two main contributions for this second model. Last, we turn our attention to perfect secrecy generation for fixed signal observation lengths as well as for their asymptotic limits. In contrast with the analysis of the previous two models that relies on probabilistic techniques, perfect secret key generation bears the essence of "zero-error information theory," and accordingly, we rely on mathematical techniques of a combinatorial nature. The model under consideration is the "Pairwise Independent Network" (PIN) model in which every pair of terminals share a random binary string, with the strings shared by distinct pairs of terminals being mutually independent. This model, which is motivated by practical aspects of a wireless communication network in which terminals communicate on the same frequency, results in three main contributions. First, the concept of perfect omniscience in data compression leads to a single-letter formula for the perfect secret key capacity of the PIN model; moreover, this capacity is shown to be achieved by linear noninteractive public communication, and coincides with strong secret key capacity. Second, taking advantage of a multigraph representation of the PIN model, we put forth an efficient algorithm for perfect secret key generation based on a combinatorial concept of maximal packing of Steiner trees of the multigraph. When all the terminals seek to share perfect secrecy, the algorithm is shown to achieve capacity. When only a subset of terminals wish to share perfect secrecy, the algorithm is shown to achieve at least half of it. Additionally, we obtain nonasymptotic and asymptotic bounds on the size and rate of the best perfect secret key generated by the algorithm. These bounds are of independent interest from a purely graph theoretic viewpoint as they constitute new estimates for the maximum size and rate of Steiner tree packing of a given multigraph. Third, a particular configuration of the PIN model arises when a lone "helper" terminal aids all the other "user" terminals generate perfect secrecy. This model has special features that enable us to obtain necessary and sufficient conditions for Steiner tree packing to achieve perfect secret key capacity.
Wyner's work on wiretap channels and the recent works on information theoretic security are based on random codes. Achieving information theoretical security with practical coding schemes is of definite interest. In this note, the attempt is to overcome this elusive task by employing the polar coding technique of Arikan. It is shown that polar codes achieve nontrivial perfect secrecy rates for binary-input degraded wiretap channels while enjoying their low encoding-decoding complexity. In the special case of symmetric main and eavesdropper channels, this coding technique achieves the secrecy capacity. Next, fading erasure wiretap channels are considered and a secret key agreement scheme is proposed, which requires only the statistical knowledge of the eavesdropper channel state information (CSI). The enabling factor is the creation of advantage over Eve, by blindly using the proposed scheme over each fading block, which is then exploited with privacy amplification techniques to generate secret keys.
We investigate the possibility of covert and secret key generation over a discrete memoryless channel model with one way public discussion. Protocols are required to conceal not only the key but also whether a protocol is being implemented. For some models, we show that covert secret key generation is possible and characterize the covert secret key capacity in special cases; in particular, the covert secret key capacity is sometimes equal to the covert capacity of the channel, so that secrecy comes “for free.” Our main contribution is the analysis of a protocol that exploits the likelihood encoder to circumvent source coding with side information and privacy amplification.
For a discrete or a continuous source model, we study the problem of secret-key generation with one round of rate-limited public communication between two legitimate users. Although we do not provide new bounds on the wiretap secret-key (WSK) capacity for the discrete source model, we use an alternative achievability scheme that may be useful for practical applications. As a side result, we conveniently extend known bounds to the case of a continuous source model. Specifically, we consider a sequential key-generation strategy, that implements a rate-limited reconciliation step to handle reliability, followed by a privacy amplification step performed with extractors to handle secrecy. We prove that such a sequential strategy achieves the best known bounds for the rate-limited WSK capacity (under the assumption of degraded sources in the case of two-way communication). However, we show that, unlike the case of rate-unlimited public communication, achieving the reconciliation capacity in a sequential strategy does not necessarily lead to achieving the best known bounds for the WSK capacity. Consequently, reliability and secrecy can be treated successively but not independently, thereby exhibiting a limitation of sequential strategies for rate-limited public communication. Nevertheless, we provide scenarios for which reliability and secrecy can be treated successively and independently, such as the two-way rate-limited SK capacity, the one-way rate-limited WSK capacity for degraded binary symmetric sources, and the one-way rate-limited WSK capacity for Gaussian degraded sources.
In this chapter, we discuss the construction of practical codes for secrecy. The design of codes for the wiretap channel turns out to be surprisingly difficult, and this area of information-theoretic security is still largely in its infancy. To some extent, the major obstacles in the road to secrecy capacity are similar to those that lay in the path to channel capacity: the random-coding arguments used to establish the secrecy capacity do not provide explicit code constructions. However, the design of wiretap codes is further impaired by the absence of a simple metric, such as a bit error rate, which could be evaluated numerically. Unlike codes designed for reliable communication, whose performance is eventually assessed by plotting a bit-error-rate curve, we cannot simulate an eavesdropper with unlimited computational power; hence, wiretap codes must possess enough structure to be provably secure. For certain channels, such as binary erasure wiretap channels, the information-theoretic secrecy constraint can be recast in terms of an algebraic property for a code-generator matrix. Most of the chapter focuses on such cases since this algebraic view of secrecy simplifies the analysis considerably. As seen in Chapter 4, the design of secret-key distillation strategies is a somewhat easier problem insofar as reliability and security can be handled separately by means of information reconciliation and privacy amplification. Essentially, the construction of coding schemes for key agreement reduces to the design of Slepian–Wolf-like codes for information reconciliation, which can be done efficiently with low-density parity-check (LDPC) codes or turbo-codes.