Sum Secrecy Rate Maximization for Full-Duplex Two-Way Relay Networks Using Alamouti-Based Rank-Two Beamforming

Qiang Li, Wing-Kin Ma, Dong Han

TL;DR

This paper tackles the challenge of maximizing secure communication rates in full-duplex two-way relay networks, where two legitimate users exchange information via a relay while an eavesdropper attempts to intercept it. The authors propose a novel optimization framework using semidefinite relaxation and minorization-maximization techniques, offering an efficient and guaranteed approach to achieve secure communication.

What Problem Does This Paper Address?

In modern wireless communication, the demand for both high data rates and robust security is paramount. Full-duplex (FD) technology, which allows simultaneous transmission and reception on the same frequency band, promises to double spectral efficiency and opens new avenues for physical-layer security. This paper focuses on a specific FD scenario: a two-way relay network where two users (Alice and Bob) communicate through an FD relay, and a single-antenna eavesdropper (Eve) tries to listen in. The core problem is to design the relay's transmission strategy – specifically, its beamforming and artificial noise (AN) generation – to maximize the sum secrecy rate (the combined secure data rate for Alice and Bob) while respecting power constraints. Existing work has explored FD security, but this paper introduces a more general formulation that includes direct links between legitimate users and the eavesdropper, and a specific Alamouti-based rank-two beamforming strategy for the relay.

Methods

The paper's main contribution lies in its optimization approach to the Sum Secrecy Rate Maximization (SSRM) problem. This problem is inherently non-convex, making it difficult to solve directly. The authors employ a two-pronged strategy:

  1. Semidefinite Relaxation (SDR) and Minorization-Maximization (MM): They first transform the problem into a form amenable to SDR. SDR is a powerful technique that relaxes a non-convex problem into a convex one by replacing a rank constraint with a positive semidefinite constraint. This relaxed problem is then tackled using an MM algorithm. The MM approach iteratively finds a lower bound for the objective function and maximizes this bound, guaranteeing convergence to a stationary solution of the relaxed problem. A key theoretical contribution is proving that the SDR is "tight" for this specific problem, meaning the solution to the relaxed problem is also a solution to the original non-convex problem.
  2. Inexact MM Approach: Recognizing that solving the convex subproblems in the exact MM algorithm can be computationally intensive, the authors propose an "inexact" MM approach. Instead of finding the exact optimal solution to each subproblem, they use an iteration-limited projected gradient method. This significantly reduces computational complexity while still guaranteeing convergence to a stationary solution of the original SSRM problem.

The paper also extends these methods to a scenario where the eavesdropper is an energy-harvesting user, adding an energy constraint for Eve.

Datasets

Not discussed in the paper. The evaluation is based on simulations with randomly generated channel models.

Key Insights and Ideas

  • FD for Security: Full-duplex technology isn't just about speed; it can be a powerful tool for enhancing physical-layer security by allowing the relay to simultaneously transmit information and generate artificial noise to jam eavesdroppers.
  • SDR Tightness is Key: Proving that the semidefinite relaxation is "tight" for this specific problem is crucial. It means the computationally tractable relaxed problem yields the same optimal solution as the original, intractable non-convex problem.
  • Inexact MM for Practicality: The proposed inexact MM algorithm offers a significant practical advantage. By trading off a small amount of optimality in each subproblem for computational speed, it makes the solution feasible for real-world applications.
  • Dual Role of AN: In the energy-harvesting eavesdropper scenario, artificial noise plays a dual role: jamming Eve and, potentially, providing her with energy. This adds an interesting layer of complexity to the optimization.
  • Trade-off Between Iterations and Complexity: The inexact MM approach highlights a trade-off: more inner-loop gradient projection steps (higher $L_k$) lead to fewer outer-loop MM iterations but increase the complexity of each MM iteration. Finding the right balance is key for efficiency.

Key Results

  • Guaranteed Convergence: Both the exact and inexact MM algorithms are proven to converge to a stationary solution of the SSRM problem.
  • Efficiency of Inexact MM: Simulations show that the inexact MM approach is significantly faster than the exact MM approach, especially when using a variable number of inner-loop iterations.
  • Performance Gains: The proposed full-duplex (FD) designs generally achieve higher sum secrecy rates compared to half-duplex (HD) designs, particularly when self-interference is well-managed.
  • Impact of Parameters: The paper demonstrates how parameters like source power, residual self-interference ($ \kappa $), and the number of relay antennas affect the sum secrecy rate. For instance, higher source power initially boosts secrecy but can degrade it beyond a certain point due to increased self-interference.
  • Energy Harvesting Eve: The framework is extended to handle an energy-harvesting eavesdropper, showing that the system can still achieve secure communication while satisfying Eve's energy needs to some extent.

Limitations and Future Work

  • Channel State Information (CSI): The proposed algorithms assume perfect CSI. Future work could explore robust designs that are less sensitive to CSI errors.
  • Complexity of Exact MM: While the inexact MM is efficient, the exact MM subproblem still requires solving a convex optimization problem, which can be computationally demanding.
  • Practical Implementation: The paper focuses on theoretical guarantees and simulation results. Further research could investigate practical implementation challenges, such as hardware imperfections and real-time CSI acquisition.
  • More Complex Scenarios: Extensions to more complex network topologies, multiple eavesdroppers, or different FD relaying strategies could be explored.

Summary updated

Abstract

Consider a two-way communication scenario where two single-antenna nodes, operating under full-duplex mode, exchange information to one another through the aid of a (full-duplex) multi-antenna relay, and there is another single-antenna node who intends to eavesdrop. The relay employs artificial noise (AN) to interfere the eavesdropper's channel, and amplify-forward (AF) Alamouti-based rank-two beamforming to establish the two-way communication links of the legitimate nodes. Our problem is to optimize the rank-two beamformer and AN covariance for sum secrecy rate maximization (SSRM). This SSRM problem is nonconvex, and we develop an efficient solution approach using semidefinite relaxation (SDR) and minorization-maximization (MM). We prove that SDR is tight for the SSRM problem and thus introduces no loss. Also, we consider an inexact MM method where an approximately but computationally cheap MM solution update is used in place of the exact update in conventional MM. We show that this inexact MM method guarantees convergence to a stationary solution to the SSRM problem. The effectiveness of our proposed approach is further demonstrated by an energy-harvesting scenario extension, and by extensive simulation results.
BibTeX
@article{li2016sum,
  author = {Qiang Li and Wing-Kin Ma and Dong Han},
  title = {{Sum Secrecy Rate Maximization for Full-Duplex Two-Way Relay Networks Using Alamouti-Based Rank-Two Beamforming}},
  journal = {IEEE JSTSP},
  year = {2016},
  doi = {10.1109/jstsp.2016.2603970},
}
Figure