Power allocation for non-orthogonal artificial noise in small-scale MIMO secure communication systems

Abstract

Physical layer security is essential for wireless communications, with artificial noise (AN) being a key technique to enhance confidentiality. However, conventional AN suffers from strict orthogonality constraints and limited design flexibility under rank-deficient channel conditions. This paper investigates power allocation for non-orthogonal artificial noise (SCO-AN), which operates in the range space of the legitimate channel, offering greater design freedom and improved secrecy. We formulate the power allocation problem between the information-bearing signal and SCO-AN under perfect channel state information (CSI), imperfect CSI, and multi-user interference. Due to the non-convex nature of the secrecy capacity function, sequential quadratic programming (SQP) is employed to obtain suboptimal power allocation solutions. Simulation results demonstrate that the proposed SQP-based scheme significantly enhances secrecy capacity compared to baseline methods, achieving up to 15% improvement in SISO systems and 20% in MIMO configurations. Computational complexity analysis confirms the efficiency of the SQP approach, requiring fewer iterations and lower per-iteration operations than alternative optimization algorithms. However, the primary contribution of this work lies in the problem formulation for SCO-AN, while the optimization method itself is a standard technique; moreover, practical factors such as channel estimation errors and hardware constraints are not fully explored, and comparisons with recent learning-based approaches as well as scalability analysis for larger MIMO systems are left for future work.