Multi-objective Aerial Collaborative Secure Communication Optimization via Generative Diffusion Model-enabled Deep Reinforcement Learning

TL;DR

This work tackles secure communications for a UAV swarm forming a virtual antenna array to counter mobile eavesdroppers. It formulates ASCEE-MOP to maximize secrecy rate while minimizing flight energy and introduces GDMTD3, a generative diffusion model-enabled TD3 framework that casts the problem as an MDP and samples actions via diffusion modeling. The approach demonstrates superior secrecy-rate performance and lower energy consumption compared with deployment policies and standard DRL benchmarks, and shows robustness to parameter variations and different UAV counts. The method advances practical secure aerial networking by combining deep reinforcement learning with generative diffusion to better navigate high-dimensional, dynamic action spaces. The results suggest meaningful implications for energy-efficient, secure UAV communications in 6G and beyond.

Abstract

Due to flexibility and low-cost, unmanned aerial vehicles (UAVs) are increasingly crucial for enhancing coverage and functionality of wireless networks. However, incorporating UAVs into next-generation wireless communication systems poses significant challenges, particularly in sustaining high-rate and long-range secure communications against eavesdropping attacks. In this work, we consider a UAV swarm-enabled secure surveillance network system, where a UAV swarm forms a virtual antenna array to transmit sensitive surveillance data to a remote base station (RBS) via collaborative beamforming (CB) so as to resist mobile eavesdroppers. Specifically, we formulate an aerial secure communication and energy efficiency multi-objective optimization problem (ASCEE-MOP) to maximize the secrecy rate of the system and to minimize the flight energy consumption of the UAV swarm. To address the non-convex, NP-hard and dynamic ASCEE-MOP, we propose a generative diffusion model-enabled twin delayed deep deterministic policy gradient (GDMTD3) method. Specifically, GDMTD3 leverages an innovative application of diffusion models to determine optimal excitation current weights and position decisions of UAVs. The diffusion models can better capture the complex dynamics and the trade-off of the ASCEE-MOP, thereby yielding promising solutions. Simulation results highlight the superior performance of the proposed approach compared with traditional deployment strategies and some other deep reinforcement learning (DRL) benchmarks. Moreover, performance analysis under various parameter settings of GDMTD3 and different numbers of UAVs verifies the robustness of the proposed approach.

Figures (11)

• Figure 1: A UAV swarm-enabled secure surveillance network system, where a UAV swarm is deployed for surveillance tasks, transmitting sensitive data to a RBS. The security of system is challenged by a mobile eavesdropper, depicted by red dashed lines, attempting to intercept the data via wiretap links over various time slots.
• Figure 2: Schematic of GDMTD3 framework, where the generative diffusion model is integrated into the actor network of TD3 algorithm to capture complex state features and generate optimal actions according to the current state of the environment.
• Figure 3: The diffusion-enabled actor network architecture, where Mish activation function Misra2020 is adopted.
• Figure 4: Comparison results of the proposed GDM-enabled DRL approach and other four deployment policies. (a) Average secrecy rate per step. (b) Average flight energy consumption per step.
• Figure 5: Comparison results of GDMTD3 and DRL benchmarks. (a) Reward per episode. (b) Average secrecy rate per step. (c) Average flight energy consumption per step.