.png){kind=icon}
.png){kind=icon}
Return
DDPG Optimization for Hybrid RIS-Assisted Secure Multi-User-MISO Networks
.png){kind=icon}
OA
AI
DOI：10.1109/OJCOMS.2026.3675029.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper investigates a Deep Reinforcement Learning (DRL) based optimization framework for secure communications in hybrid reconfigurable intelligent surface (RIS)-assisted multi-user multiple-input single-output (MU-MISO) systems. A secrecy-aware actor-critic model is developed using deep deterministic policy gradient (DDPG) to optimize the hybrid RIS reflection and amplification coefficients with a fixed Zero-Forcing transmit beamformer under secrecy constraints. A novel reward design is proposed that incorporates both user fairness and secrecy rate maximization, penalizing imbalance and eavesdropper leakage. Unlike conventional DRL-based RIS designs that maximize only aggregate secrecy rate, the proposed fairness-aware secrecy reward explicitly regulates the standard deviation of per-user secrecy rates to promote balanced performance across users. The channel model captures both direct and reflected links, with independent Rayleigh fading for all paths. The simulation results reveal that the proposed approach significantly improves the average sum secrecy rate and user throughput as transmit power and RIS size increase. The results also demonstrate that the hybrid RIS outperforms the passive RIS and achieves secrecy performance comparable to that of an active RIS with reduced power consumption. Furthermore, the proposed DDPG-based method achieves secure and efficient resource allocation, making it a promising candidate for next-generation secure RIS-assisted wireless networks.
Keywords:
Reconfigurable intelligent surface (RIS)
hybrid RIS
deep reinforcement learning
deep deterministic policy gradient (DDPG)
MISO
beamforming
Journal
I
IF:
6.1
Papers:
489
Citations:
0
View more organizations