AI-Driven Aerial Relay Placement and Power Allocation for Enhanced Secrecy in Untrusted UAV-Enabled Networks

In this paper, we introduce an innovative artificial intelligence (AI)-driven methodology for optimizing the deployment of unmanned aerial vehicle (UAV)-enabled networks, specifically addressing scenarios where an untrusted UAV serves as an amplify-and-forward (AF) relay while potentially engaging in eavesdropping activities. The primary goal is to enhance secrecy performance by optimizing both the three-dimensional (3-D) positioning of the UAV relay and the power allocation strategy. Conventional optimization methods often fail to effectively capture the intricate, dynamic nature of such networks, thereby limiting their ability to ensure adaptive secrecy optimization. To address this challenge, we propose a hybrid approach that integrates particle swarm optimization (PSO) with the K-means and K-medoids unsupervised machine learning (ML) techniques. This combined approach not only refines UAV placement but also ensures adherence to the derived rule-based power allocation conditions, which are crucial for maximizing secrecy performance. Through extensive simulations, we evaluate the efficacy of the proposed K-PSO methodology against standalone PSO, K-means, and K-medoids models. The results demonstrate that the hybrid K-PSO approach consistently outperforms its standalone counterparts by significantly improving the average achievable secrecy rate while reducing the secrecy outage probability (SOP).