Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Shadow Wireless Intelligence: Large Language Model-Driven Reasoning in Covert Communications

Yuanai Xie, Zhaozhi Liu, Xiao Zhang, Shihua Zhang, Rui Hou, Minrui Xu, Ruichen Zhang, Dusit Niyato

TL;DR

This paper presents Shadow Wireless Intelligence (SWI), a framework that integrates Large Language Model (LLM) reasoning with retrieval-augmented generation to enable real-time, context-aware decision-making for covert wireless communications in 6G networks. SWI combines a mixture-of-experts LLM (DeepSeek-R1) with a CC-specific knowledge base, chain-of-thought prompting, and retrieval augmentation to derive optimized covert strategies such as artificial noise power while respecting covertness constraints, demonstrated through a full-duplex CC case study. The results show that DeepSeek-R1 achieves 85% accuracy in deriving closed-form expressions and 94% correctness in simulated code generation, outperforming baseline models and highlighting SWI’s robustness, interpretability, and adaptability in adversarial wireless environments. The work lays a foundation for LLM-driven secure wireless systems and suggests future extensions to multi-user scenarios, graph-based retrieval, and distributed reasoning to scale real-time intelligent covert optimization in 6G networks.

Abstract

Covert Communications (CC) can secure sensitive transmissions in industrial, military, and mission-critical applications within 6G wireless networks. However, traditional optimization methods based on Artificial Noise (AN), power control, and channel manipulation might not adapt to dynamic and adversarial environments due to the high dimensionality, nonlinearity, and stringent real-time covertness requirements. To bridge this gap, we introduce Shadow Wireless Intelligence (SWI), which integrates the reasoning capabilities of Large Language Models (LLMs) with retrieval-augmented generation to enable intelligent decision-making in covert wireless systems. Specifically, we utilize DeepSeek-R1, a mixture-of-experts-based LLM with RL-enhanced reasoning, combined with real-time retrieval of domain-specific knowledge to improve context accuracy and mitigate hallucinations. Our approach develops a structured CC knowledge base, supports context-aware retrieval, and performs semantic optimization, allowing LLMs to generate and adapt CC strategies in real time. In a case study on optimizing AN power in a full-duplex CC scenario, DeepSeek-R1 achieves 85% symbolic derivation accuracy and 94% correctness in the generation of simulation code, outperforming baseline models. These results validate SWI as a robust, interpretable, and adaptive foundation for LLM-driven intelligent covert wireless systems in 6G networks.

Shadow Wireless Intelligence: Large Language Model-Driven Reasoning in Covert Communications

TL;DR

This paper presents Shadow Wireless Intelligence (SWI), a framework that integrates Large Language Model (LLM) reasoning with retrieval-augmented generation to enable real-time, context-aware decision-making for covert wireless communications in 6G networks. SWI combines a mixture-of-experts LLM (DeepSeek-R1) with a CC-specific knowledge base, chain-of-thought prompting, and retrieval augmentation to derive optimized covert strategies such as artificial noise power while respecting covertness constraints, demonstrated through a full-duplex CC case study. The results show that DeepSeek-R1 achieves 85% accuracy in deriving closed-form expressions and 94% correctness in simulated code generation, outperforming baseline models and highlighting SWI’s robustness, interpretability, and adaptability in adversarial wireless environments. The work lays a foundation for LLM-driven secure wireless systems and suggests future extensions to multi-user scenarios, graph-based retrieval, and distributed reasoning to scale real-time intelligent covert optimization in 6G networks.

Abstract

Covert Communications (CC) can secure sensitive transmissions in industrial, military, and mission-critical applications within 6G wireless networks. However, traditional optimization methods based on Artificial Noise (AN), power control, and channel manipulation might not adapt to dynamic and adversarial environments due to the high dimensionality, nonlinearity, and stringent real-time covertness requirements. To bridge this gap, we introduce Shadow Wireless Intelligence (SWI), which integrates the reasoning capabilities of Large Language Models (LLMs) with retrieval-augmented generation to enable intelligent decision-making in covert wireless systems. Specifically, we utilize DeepSeek-R1, a mixture-of-experts-based LLM with RL-enhanced reasoning, combined with real-time retrieval of domain-specific knowledge to improve context accuracy and mitigate hallucinations. Our approach develops a structured CC knowledge base, supports context-aware retrieval, and performs semantic optimization, allowing LLMs to generate and adapt CC strategies in real time. In a case study on optimizing AN power in a full-duplex CC scenario, DeepSeek-R1 achieves 85% symbolic derivation accuracy and 94% correctness in the generation of simulation code, outperforming baseline models. These results validate SWI as a robust, interpretable, and adaptive foundation for LLM-driven intelligent covert wireless systems in 6G networks.
Paper Structure (14 sections, 5 figures)

This paper contains 14 sections, 5 figures.

Table of Contents

  1. Introduction
  2. OVERVIEW OF LLM-DRIVEN REASONING IN CC
  3. Covert Communications
  4. LLMs for CC
  5. CC for LLMs
  6. SWI Reasoning in CC
  7. Shadow Wireless Intelligence Framework
  8. CASE STUDY: LLM-Driven Reasoning in CC
  9. Optimal Power Derivation Experiment
  10. Problem Description
  11. Experimental Settings
  12. Problem-Solving Process
  13. Experimental Results
  14. Conclusions and Future Directions

Figures (5)

  • Figure 1: Overview of CC technologies: scenarios, evolution, optimization challenges, and enhancement strategies.
  • Figure 2: Advantages and benefits: LLM in CC-based optimization and CC for LLM inference.
  • Figure 3: SWI outperforms LLM-only by combining CoT reasoning and external data.
  • Figure 4: Workflow of SWI for optimal noise power under covertness constraints.
  • Figure 5: Comparison of Model Advantages, Disadvantages, and Reasoning Precision in the Case Study.