Resumen de Polar coding for the wiretap broadcast channel
In the next era of communications, where heterogeneous, asynchronous and ultra-low latency networks are drawn on the horizon, classical cryptography might be inadequate due to the excessive cost of maintaining a public-key infrastructure and the high computational capacity required in the devices. Moreover, it is becoming increasingly difficult to guarantee that the computational capacity of adversaries would not be able to break the cryptograms. Consequently, information-theoretic security might play an important role in the future development of these systems. The notion of secrecy in this case does not rely on any assumption of the computational power of eavesdroppers, and is based instead on guaranteeing statistical independence between the information message and the observed cryptogram. This is possible by constructing channel codes that exploit the noisy behavior of the channels involved in the communication.
Although there has been very substantial research in the last two decades regarding information-theoretic security, little has gone to study and design practical codes for keyless secret communication. In recent years, polar codes have changed the lay of the land because they are the first constructive and provable channel codes that are able to provide reliability and information-theoretic secrecy simultaneously. Additionally, their explicit construction and the low complexity of the encoding/decoding schemes makes them suitable for the new generation of communication systems.
The main objective of this dissertation is to provide polar coding schemes that achieve the best known inner-bounds on the capacity regions of different multiuser models over the discrete memoryless broadcast channel. These models not only impose a reliability constraint, but also some sort of information-theoretic secrecy condition in the presence of eavesdroppers. In general, we focus on describing the construction and the encoding/decoding schemes of the the proposed polar code for a particular setting. Then, we analyze the reliability and the secrecy performance of these schemes in order to prove that they are able to achieve these inner-bounds as the blocklength tends to infinity.
The first part of the thesis drives the attention to two different models over the degraded broadcast channel that commonly appear in real communication systems. In this models, there are a set of legitimate receivers and a set of eavesdroppers that can be ordered based on the quality of their channels. According to this ordering, different reliability or secrecy constraints apply for each legitimate receiver or eavesdropper respectively. Moreover, we propose practical methods for constructing the polar codes for both models and analyze the performance of the coding schemes by means of simulations. Despite we only evaluate the construction for these two particular settings, the proposed methods are also suitable for any polar coding scheme that must satisfy some reliability and secrecy conditions simultaneously.
In the second part of the dissertation we describe and analyze two different polar coding schemes for the general broadcast channel (where channels are not necessarily degraded) with two legitimate receivers and one eavesdropper. We consider a model where a confidential and a non-confidential message must be reliably decoded by both legitimate receivers in presence of an eavesdropper. Despite it is almost immediate to find an inner-bound on the capacity for this model using random coding arguments, how to secretly convey the same confidential message to both legitimate receivers using polar codes is not straightforward. We also analyze the setting where a transmitter wants to send different confidential and non-confidential messages to the corresponding legitimate receivers. We compare two inner-bounds on the capacity of this model, and we design a polar coding scheme that achieves the inner-bound that surely includes the other.
