Deep-Learning-Based Physical-Layer Secret Key Generation for FDD Systems

In cellular systems, deploying a large number of antennas at the base station (BS), also called massive multiple-input multiple-output (MIMO), can offer a huge improvement in system throughput. This performance gain achieved through coherent beamforming highly depends on having accurate channel state information (CSI) at the BS. Initially, massive MIMO was considered promising only for time division duplexing (TDD) systems because the downlink training overhead in frequency division duplexing (FDD) systems could be large. To reduce the overhead in FDD systems, we propose efficient eigenspace training and precoding. In the downlink training, the prebeamforming matrix is designed based on the spatial fading correlation to probe the channel and minimize the mean squared error (MSE) in channel estimation. In data transmission, the precoding matrix is designed based on instantaneous CSI to manage the inter-user interference. Our algorithms achieve significant savings in the downlink training because the overhead is associated with the dimension of prebeamforming matrix and no longer limited by the number of BS antennas. Compared with joint spatial division and multiplexing, our algorithms offer much lower MSE and, under some conditions, higher sum rates.

Channel reciprocity is limited in frequency division duplex (FDD) system, and thus those channel reciprocity-based secret key extraction methods which are widely adopted in time division duplex (TDD) systems can hardly be implemented in FDD systems. In this paper, a novel secret key extraction method in OFDM-FDD system is proposed to achieve physical-layer security, by which the channel frequency response (CFR) in the uplink at the same time is estimated by the communication pair with the aid of a specially designed forwarding strategy based on confidential probing signal, and then the secret key is extracted from the CFR estimates. The effectiveness of the proposed method is verified by numerical simulations.

Massive MIMO is a powerful multiuser/multiantenna technology that exploits a very large number of antennas at the base station side and the knowledge of the channel matrix between base station antennas and multiple users in order to achieve large beamforming and multiplexing gain. Classical massive MIMO exploits Time-Division Duplexing (TDD) and channel reciprocity, such that the channel matrix can be learned at the base station from the incoming uplink pilot signals sent by the users. However, the large majority of cellular networks deployed today make use of Frequency Division Duplexing (FDD) where channel reciprocity does not hold and explicit downlink probing and uplink CSI feedback are required in order to achieve some spatial multiplexing gain. Unfortunately, the overhead incurred by explicit probing and feedback is very large in massive MIMO, since the channels are high-dimensional random vectors. In this paper, we present a new approach to achieve very competitive tradeoff between spatial multiplexing gain and probing/feedback overhead in FDD massive MIMO. Our approach is based on two novel concepts: 1) an efficient and mathematically rigorous technique to extrapolate the channel covariance matrix from the uplink to the downlink, such that the second order statistics of each downlink channel can be accurately learned for free from uplink pilots; 2) a novel “sparsifying precoding” approach, that introduces sparsity in the channel in a controlled form, such that for any assigned overhead (i.e., downlink pilot dimension) it is possible to set an optimal sparsity level for which the “effective” channels after sparsification can be estimated at the base station with low mean-square error. We compare our method with that of the state-of-the-art compressed sensing (CS) based method. Our results show that the proposed method is much more robust than compressed sensing methods, since it is able to “shape the channel sparsity” as desired, instead of being at the mercy of nature (i.e., at the mercy of the natural sparsity induced by the nronaaation environment).

Physical layer security (PLS) is considered one of the most promising solutions to solve the security problems of massive Internet of Things (IoTs) devices because of its lightweight and high efficiency.Significantly, the recent physical layer key generation (PLKG) scheme based on transmission delay proposed by Huang et al. does not have any restrictions on communication methods and can extend the traditional physical layer security based on wireless channels to the whole Internet scene.However, the secret-sharing strategy adopted in this scheme has hidden dangers of collusion attack, which may lead to security problems such as information tampering and privacy disclosure.By establishing a probability model, this paper quantitatively analyzes the relationship between the number of malicious collusion nodes and the probability of key exposure, which proves the existence of this security problem.In order to solve the problem of collusion attack in Huang et al.'s scheme, this paper proposes an anti-collusion attack defense method, which minimizes the influence of collusion attack on key security by optimizing parameters including the number of the middle forwarding nodes, the random forwarding times, the time delay measurement times and the out-of-control rate of forwarding nodes.Finally, based on the game model, we prove that the defense method proposed in this paper can reduce the risk of key leakage to zero under the scenario of the "Careless Defender" and "Cautious Defender" respectively.

Key generation rate and bit disagreement ratio are two main indicators to evaluate the performance of physical layer key generation (PLKG). This work explores the key generation performance in reconfigurable intelligent surface (RIS)-assisted PLKG, where a concrete scheme of key generation is proposed. Specifically, we first deduce the key generation rate expression based on estimation theory in RIS-assisted PLKG. Then we optimize the RIS reflecting coefficients with instantaneous channel state information to maximize the key generation rate. Finally, aiming at minimizing transmit power while guaranteeing key generation rate target, we formulate a power minimization problem in RIS-assisted PLKG. An alternating optimization algorithm is applied to solve the non-convex mixed-integer nonlinear programming. Simulation results demonstrate that the key generation rate of the proposed scheme is up to 197.5% higher than the existing relay-assisted scheme, and we successfully reduce the required transmit power by at least 6.6 dBm than without power minimization scheme.

A novel method for performing predictive closed-loop power control is used to illustrate the predictability of uplink channels from downlink channels in a frequency division duplex (FDD) wireless link. In an uplink power control scenario, the proposed scheme performs uplink prediction using downlink-derived (linear) predictor coefficients. Essentially, the lack of correlation between the fading processes on uplink and downlink (due to FDD and frequency selective fading) is circumvented by using the uplink fading history, together with a set of link-independent, autoregressive (AR) coefficients. An AR model is used to formally describe the frequency selective fading process on either link, regardless of the number of Rayleigh paths. It is also proven that the relevant AR coefficients are common to both links, up to a proper normalization, irrespective of whether the number of paths on the uplink differs from that on the downlink. Both an uncoded and a coded system are simulated, and their performance results are compared, on equal grounds, with those of the corresponding nonpredictive systems.

Spectrum and infrastructure sharing among multiple mobile network operators have recently drawn significant attention from the industry and the academia sectors. Recent advances in the area of Software Defined Networking (SDN) launch the design of innovative SDN-based solutions for resource sharing in multi-tier networks of macrocell and small-sized base stations, a.k.a. HetNet. In this paper, we present a novel SDN-based framework that enables efficient and elastic spectrum utilization among multiple operators in 3GPP LTE-A HetNet scenario. Assuming a multi-operator environment of Frequency Division Duplex (FDD) macrocells complemented by multi-tenant Time Division Duplex (TDD) pico cells, we present a SDN-based architecture that allows efficient resource sharing among the TDD and FDD systems in a dynamic way. A TDD frame re-configuration mechanism is also employed, to optimize the ratio of uplink and downlink slots in the TDD frame of picocells. System-level simulation results demonstrate that the combination of these two functional enhancements in the LTE-A HetNet, significantly reduces the application layer delay for both the FDD macrocell and TDD picocell systems, leading to highly efficient and dynamic spectrum sharing among multiple network operators.

In order to unlock the full advantages of massive multiple input multiple output (MIMO) in the downlink, channel state information (CSI) is required at the base station (BS) to optimize the beamforming matrices. In frequency division duplex (FDD) systems, full channel reciprocity does not hold, and CSI acquisition generally requires downlink pilot transmission followed by uplink feedback. Prior work proposed the end-to-end design of pilot transmission, feedback, and CSI estimation via deep learning. In this work, we introduce an enhanced end-to-end design that leverages partial uplink-downlink reciprocity and temporal correlation of the fading processes by utilizing jointly downlink and uplink pilots. The proposed method is based on a novel deep learning architecture – HyperRNN – that combines hypernetworks and recurrent neural networks (RNNs) to optimize the transfer of long-term channel features from uplink to downlink. Simulation results demonstrate that the HyperRNN achieves a lower normalized mean square error (NMSE) performance, and that it reduces requirements in terms of pilot lengths.

Secret key generation by extracting the shared randomness in the wireless fading channel from physical layer is an interesting topic of practical value. Previous works have focused on the study of physical layer key generation with two nodes from the view point of key generation rate (KGR). Information theoretic limits and the KGRs in implementation have been derived. However, in real-world applications, the physical layer key generation problem involving multiple nodes is the common case, which lacks sufficient study so far. Multi-node case differs from two-node case in that there are two more important considerations: (1) the trade-off between KGR and probing efficiency at individual node pair; (2) channel probing schedule among multiple node pairs. This paper aims at minimizing the Overall Waiting Time of physical layer key generation with multiple users (shorten as OWT) through the optimization of probing rates at individual node pair and channel probing schedule. The theoretical lower bound of OWT is derived first, then a practical method (MUKEM) is proposed to compute reasonable probing rates and channel probing schedule for multiple node pairs to obtain a short OWT. Simulations are conducted to evaluate the effectiveness of our method. The results show that 70 % of OWT can be reduced by using our method comparing with one-by-one key generations; while it is only about 8 % longer than the lower bound of OWT.

Key generation leveraging wireless channel reciprocity can establish secret keys from unauthenticated broadcast channels and thus protect the communication of wireless networks. However, in real environments, the measurements collected by the involved transceivers might have many discrepancies due to the non-simultaneous directional measurements and other factors, such as hardware differences and asymmetric interferences. Meanwhile, if the multi-level quantization in conjunction with source coding is implemented to convert measurements to bits, the resulting high bit mismatch rate makes key agreement inefficient and insecure. In this paper, key generation schemes leveraging channel reciprocity and balanced gray code are proposed. To enhance the efficiency of key generation, an efficient encoding scheme based on balanced gray code is carried out during key generation. The proposed encoding scheme can simultaneously reduce the bit mismatch rate and improve the secret bit rate. In addition, the code book is reusable once it is generated. The proposed key generation schemes can be classified into the uniform and non-uniform schemes according to different quantization processes. Several experiments in real mobile environments are conducted to validate the proposed schemes. The received signal strength measurements are extracted from heterogeneous devices as statistics of channel states in each experiment. The results of these experiments show that (1) the proposed encoding scheme significantly enhances the efficiency of key generation; (2) both the uniform and non-uniform schemes can efficiently generate secret keys between heterogeneous transceivers.

In this work, we propose a Gaussian mixture model (GMM)-based pilot design scheme for downlink (DL) channel estimation in single- and multi-user multiple-input multiple-output (MIMO) frequency division duplex (FDD) systems. In an initial offline phase, the GMM captures prior information on the channel statistics through training, which is then utilized for pilot design. In the single-user case, the GMM is utilized to construct a codebook of pilot matrices and, once shared with the mobile terminal (MT), can be employed to determine a feedback index at the MT. This index selects a pilot matrix from the constructed codebook, eliminating the need for online pilot optimization. We further establish a sum conditional mutual information (CMI)-based pilot optimization framework for multi-user MIMO (MU-MIMO) systems. Based on the established framework, we utilize the GMM for pilot matrix design in MU-MIMO systems. The analytic representation of the GMM enables the adaptation to any signal-to-noise ratio (SNR) level and pilot configuration without re-training. Additionally, an adaption to any number of MTs is facilitated. Extensive simulations demonstrate the superior performance of the proposed pilot design scheme compared to state-of-the-art approaches. The performance gains can be exploited, e.g., to deploy systems with fewer pilots.

Shared secret key generation from wireless channel characteristics has attracted a lot of research attention in recent years due to its low computational overhead and potentials to achieve information-theoretic security. In this paper, for millimeter wave (mmWave) Massive MIMO channel, we propose to use new channel characteristics – virtual AoA (Angle of Arrival) and AoD (Angle of Departure), to generate a shared secret key between two devices. Other than reciprocity, AoA and AoD of the virtual channel present sparsity and their estimation is efficient and robust against noise. Through theoretical analysis and extensive simulations, we show that the proposed key generation using virtual AoA and AoD achieves much higher bit agreement ratio than existing mechanisms directly using channel state information. Our method can achieve above 99% bit agreement ratio even under very low SNR (e.g., −10dB), which is not achievable by the existing mechanisms. Besides, in our proposed mechanism, the bit agreement ratio becomes higher when number of antennas increases, along with higher key generate rates achieved per channel sounding.

Physical layer key generation is a promising privacy protection technique suitable for devices with limited computational ability. In this work, we design a key generation algorithm based on orthogonal frequency-division multiplexing (OFDM) waveforms and implement the algorithm on a testbed. By exploiting the holistic channel state information (CSI), we improve the key generation rate (KGR) significantly. The performance of the algorithm and the secrecy of generated keys are evaluated by experimentation. Then, we design a cross-layer encryption protocol based on our key generation algorithm and the Advanced Encryption Standard (AES). We provide a mathematical framework to show that the secrecy of the encryption is enhanced compared to traditional encryption schemes with one pre-shared key (e.g., WPA2-PSK), even when some generated keys are leaked to the eavesdropper. The results lead to practical and robust applications of physical layer key generation.

Internet-of-Things (IoT) networks generally contain resource-constrained devices that require an energy-efficient key generation procedure to producing secure keys at a faster rate. The physical characteristics of the wireless channel can be exploited to secure communication within IoT networks. In particular, secret keys can be generated by leveraging on the randomness of the wireless physical channel between two communicating parties. The conventional mechanism of generating keys at the physical layer, i.e., using channel probing, quantization, information reconciliation, and privacy amplification, may not be preferable for IoT devices. In addition, in some cases IoT devices may be deployed in static environments, wherein the channel coherence time is too high to generate keys at a faster rate and with the desired randomness. This study proposes a mapping table-based key distribution scheme for IoT environments, wherein multiple characteristics of the random channel are combined to improve not only the key generation rate (KGR) but also the key agreement rate (KAR) and bit error rate (BER). In the proposed scheme, both the channel magnitude and the phase are exploited in the key generation process. The proposed scheme is immune to channel estimation errors while providing sufficient randomness in the static environment. Additionally, the scheme is thoroughly investigated for different scenarios including the case of a smarter eavesdropper, which attempts to estimate the channel between the legitimate nodes. This case verifies the robustness of the proposed scheme in different settings and attack models.

Heterogeneous Internet of Things (IoT) and multi-access mobile edge computing (MA-MEC) are believed as supporting technologies for building a smart city. The advancement and flourish of IoT are facilitating the entry of human society into the Internet of Everything era, which lay the foundation of the smart city. To address the conflict between computation capability and low-cost mobile devices in IoT, the MA-MEC is available for supporting the resource-limited and computation-sensitive services and applications by computation offloading and distributed content delivery/caching. However, deploying cloud computing capability within the radio access network may face serious security threats, which stem from not only the existing technologies and networks but also the MA-MEC-based IoT itself. Therefore, in this paper, the solutions to address the security threats are investigated from physical layer perspectives, since physical layer security technologies have the advantages of achieving perfect secrecy, low-computational complexity, and resource consumption, and good adaptation for channel changes. Specifically, we investigate the secure wiretap coding, resource allocation, signal processing, and multi-node cooperation, along with physical layer key generation and authentication, to cope with the emerging security challenges. Finally, the paper is concluded with some possible future research directions.