Wireless Communications and Mobile Computing

Volume 2017 (2017), Article ID 5396092, 10 pages

https://doi.org/10.1155/2017/5396092

## Joint Design of Massive MIMO Precoder and Security Scheme for Multiuser Scenarios under Reciprocal Channel Conditions

^{1}Instituto de Telecomunicações and DETI, University of Aveiro, Aveiro, Portugal^{2}Instituto de Telecomunicações, Department of Electrical and Computer Engineering, University of Coimbra, Coimbra, Portugal^{3}CISUC, Department of Informatics Engineering, University of Coimbra, Coimbra, Portugal

Correspondence should be addressed to Gustavo Anjos; tp.au@sojnaovatsug

Received 25 August 2017; Revised 30 October 2017; Accepted 13 November 2017; Published 10 December 2017

Academic Editor: Daniele Pinchera

Copyright © 2017 Gustavo Anjos et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

The exploration of the physical layer characteristics of the wireless channel is currently the object of intensive research in order to develop advanced secrecy schemes that can protect information against eavesdropping attacks. Following this line of work, in this manuscript we consider a massive MIMO system and jointly design the channel precoder and security scheme. By doing that we ensure that the precoding operation does not reduce the degree of secrecy provided by the security scheme. The fundamental working principle of the proposed technique is to apply selective random rotations in the transmitted signal at the antenna level in order to achieve a compromise between legitimate and eavesdropper channel capacities. These rotations use the phase of the reciprocal wireless channel as a common random source between the transmitter and the intended receiver. To assess the security performance, the proposed joint scheme is compared with a recently proposed approach for massive MIMO systems. The results show that, with the proposed joint design, the number of antenna elements does not influence the eavesdropper channel capacity, which is proved to be equal to zero, in contrast to previous approaches.

#### 1. Introduction

The growing demand for capacity that wireless networks have experienced in recent years has resulted from the emergence of a new set of services, cheap devices, and useful applications that started to play a crucial role in the professional and social domains of people’s lives. The dependence of such services and applications increased in such a way that now they must be available anywhere, at any time, and in any circumstance. In addition to the flexible availability demand, some of these services require the exchange of private sensitive information, such as personal financial data, government-level classified information, or critical business reports. Due to the broadcast nature of the wireless channel, the protection of this kind of information in mobile networks is seen as a main system design parameter that must be carefully addressed. Since the release of the initial mobile standards, higher layer cryptographic protocols have been used as the main security platform to protect wireless communications from unintended receivers [1]. Although these protocols have found widespread acceptance, they rely on the assumption that an eavesdropper has computational resource limitations [2]. For instance, in asymmetric public key cryptosystems, the security level is supported by the assumption that the integer factorization of the product of two large prime numbers is a very intensive computational task taking into account current factorization techniques. However, in recent years, the advances in the field of number theory and the continuous increase in transistor integration levels are putting pressure on these types of protocols, forcing the use of larger key sizes, which in turn leads to a higher implementation complexity for cryptographic systems [3].

To complement the limitations of standalone cryptographic protocols [4, 5], the development of secrecy schemes that explore the physical layer characteristics of the wireless channel has been considered to efficiently improve information security in wireless networks. Physical layer security does not make any assumption regarding the level of computational capacity at the unintended receiver, being the secrecy provided by building on a channel advantage in relation to the eavesdropper [6]. The advancement of physical layer secrecy can be performed at two main levels: the coding domain and the signal level domain. In the coding domain, the target is to use error-correction codes that are designed to not only provide error detection but also implement some level of secrecy in a wiretap channel [7–10]. For signaling, techniques involving specific precoding designs, power allocation schemes, and cooperative jamming based on interference alignment (IA) [11, 12] and artificial noise injection have been defined in the literature [13–19].

The use of massive MIMO technology is being considered by the research community as mandatory evolution of the conventional MIMO systems to address the capacity requirements of future 5G mobile networks [20–23]. Over the last years, intensive research efforts have been made to solve some practical constraints associated with the large-scale deployment of massive MIMO. However, aspects related to information security have been left aside for some time, and only recently have they begun to be discussed. In [24], pilot contamination attacks in a Time Division Duplex (TDD) multicell multiuser massive MIMO scenario were analyzed. With the use of the same uplink training sequence of a legitimate receiver (Bob), the eavesdropper (Eve) can force the contamination of the channel estimated at the base station, which, in a subsequent downlink beamforming phase, will allow the unintended receiver to improve their ability to tap the communication. To address this problem, the authors derived a closed form solution for optimal power allocation between the information signal and noise, considering a maximum ratio transmission (MRT) precoder plus artificial noise (AN) generation at the legitimate transmitter. A null-space (NS) based precoder that was designed to mitigate the effect of a pilot contamination attack was also suggested. Considering again the same multicell multiuser scenario of [24], the work in [25] compared the use of NS-based precoding and random shaping matrix precoding for AN generation in an MRT-based massive MIMO transmitter under the presence of a multiantenna eavesdropper. Considering the large computational complexity required to calculate the NS of large channel matrices, the authors in [25] verified that the use of random shaping matrices for AN precoding could offer a good solution in terms of performance/complexity tradeoff. The work in [26] shows that by combining the information signal with artificial generated noise, a positive secrecy capacity can be obtained, assuming that the number of antennas at the eavesdropper is smaller than the total number of antennas at the legitimate transmitter. In the first scenario, a multiple-antenna transmitter forces the generated AN to lie in the null space of the legitimate receiver channel. In a second scenario, a single-antenna node cooperates with single-antenna relays to simulate the effect of a multiple-antenna transmitter generating AN. An attempt to force an independent relation between the secrecy capacity and the number of antennas at the eavesdropper was proposed in [27] with the development of the original symbol phase rotated (OSPR) technique. The idea of the OSPR scheme is to use the phase of the reciprocal wireless channel to define random rotations on the original data symbols that are exchanged in the downlink direction between a massive MIMO base station (BS) and several single-antenna user terminals (UTs). Considering that the reciprocal channels are available at both sides of the legitimate link, the intended receiver has all the information required to revert the original random phase rotations applied at the legitimate transmitter, while at the eavesdropper side, assuming no collocation with the legitimate UTs, the random phase rotations cannot be reverted. In [27], the authors claim that, even in the presence of a powerful massive MIMO eavesdropper equipped with an infinite number of antenna elements, the OSPR technique achieves a positive secrecy rate. Using the same basic idea considered in [27], the authors in [28] applied the OSPR scheme in the uplink direction. Another approach that exploits the channel reciprocity to provide secrecy in wireless single-antenna systems was proposed in [29]. The authors in [29] suggested a secrecy scheme that uses the reciprocal channel phase to randomly define discrete jamming signals. In the first part of the work, in order to evaluate the baseline secrecy level of the scheme, the authors consider random combinations of data and jamming signals. In the second part, an efficient data and jamming signal combining algorithm was developed, which allowed verifying a significant improvement over the secrecy level of the baseline scheme.

In this paper, we propose to jointly design the security scheme and massive MIMO precoder. The target is to provide information secrecy in a multiuser massive MIMO scenario in the presence of a passive eavesdropper equipped with a large number of antenna elements. The joint design ensures that the precoding operation does not reduce the degree of secrecy provided by the security scheme and achieves a compromise between legitimate and eavesdropper channel capacities. The fundamental working principle of the joint scheme is to create equivocation at the unintended receiver by applying antenna selective random phase rotations in both the original data symbols and the precoder. To evaluate the merit of the proposed scheme, a comparison with a technique proposed in the literature [27] was performed by using the secrecy capacity as the metric in different multiuser massive MIMO configurations. The comparison showed that for the new proposed scheme the eavesdropper channel capacity is always zero, contrary to what occurs in the scheme proposed in [27], where some leakage of information was always verified. In summary, the main contributions of the presented work are outlined in the following two points:(a)Mathematical analysis of the existing OSPR scheme using the secrecy capacity as evaluation metric: This analysis identifies some of the limitations of the OSPR scheme, which include the nonintentional phase reversions in the OSPR symbols caused by the MRT precoder that leads to zero secrecy capacity for the OSPR scheme. Moreover, the mathematical analysis is confirmed by simulation.(b)Proposal of a joint design for the massive MIMO precoder and security scheme which removes the drawback of the OSPR scheme: We show analytically and by simulation that the proposed joint design forces the capacity of the eavesdropper channel always to zero, independently of the number of antennas at the eavesdropper. Furthermore, the zero channel capacity at the eavesdropper is obtained with minimal impact in the legitimate user’s channel.

The remainder of the paper is organized as follows: Section 2 defines the general system characterization and the secrecy metrics used in the numerical evaluations. Section 3 starts by a description of the OSPR scheme proposed in [27] followed by the mathematical analysis of this secrecy scheme that enables the identification of secrecy breaches, justifying therefore the need for new approaches. The security scheme proposed in this manuscript is formulated in Section 4. In Section 5, the numerical evaluation results are presented. Finally, the main conclusions are outlined in Section 6.

*Notations.* Boldface capital letters denote matrices and boldface lowercase letters denote column vectors. The operations , , , and represent the transpose, the Hermitian transpose, the conjugate, and the trace of a matrix, respectively. Consider a vector ; corresponds to a diagonal matrix with diagonal entries equal to vector . The norm of vector is defined as .

#### 2. System Model and Metrics

In this section, the system setup, as well as the evaluation metrics used to assess the schemes performance, is presented.

##### 2.1. System Model

Figure 1 depicts the general setup used in the schemes described in Sections 3 and 4. The system is a multiuser massive MIMO cell with single-antenna user terminals (UT), one base station (BS), and one passive eavesdropper (Eve) employing and antennas, respectively. The assumption of a passive eavesdropper means that this node listens to the communication and does not cause any intentional interference in the communication channel, making his presence and location uncertain to the legitimate transmitter. In this work, Eve wants to tap the information that is exchanged between the BS and the UTs. We consider TDD channel reciprocity and perfect channel estimations at the BS, which are acquired through an uplink training process. Additionally, we assume that Eve is not collocated with any of the UTs nodes, that is, independence between all the channel responses is verified. In Figure 1, , represents the data symbol of user . All the channel responses are modeled by zero mean and unity variance complex Gaussian fading coefficients with as the channel matrix between the BS and all of the UTs, where , , , denotes the entry at row and column of matrix . In this work, ideal RF up- and downconversion are assumed with all the baseband processing applied to an independent flat fading channel realization.