Hybrid Optimization for IRS-Aided Secure Communication: A Gradient Projection and Dynamic Security Framework

The emergence of advanced wireless communication technologies has heightened the need for effective physical-layer security, particularly in multi-user environments susceptible to eavesdropping. In this work, we propose a Hybrid Gradient Descent Optimization (HGDO) framework that leverages an Intelligent Reflecting Surface (IRS) to safeguard wireless transmissions. Our methodology features a hierarchical two-phase design, a Gradient Projection Method that jointly refines beamforming vectors and IRS phase shifts, and a Dynamic Security Region Expansion (DSRE) algorithm that reacts in real time to shifts in channel conditions. Simulation results confirm that the HGDO framework achieves notable improvements in secrecy rates, energy efficiency, and adaptability under diverse system settings. For instance, with 100 IRS elements, the solution attains a secrecy rate of 42 bps/Hz, showcasing its scalability and reliability. These findings suggest that HGDO holds strong promise for fortifying next-generation wireless networks against emerging security threats.