A Trustworthy Multi-LLM Network: Challenges,Solutions, and A Use Case

TL;DR

This paper tackles trust and reliability issues in LLM-driven network optimization by proposing a blockchain-based Trustworthy Multi-LLM Network (MultiLLMN). By decentralizing consensus across multiple LLMs, the framework aims to mitigate biases, hallucinations, and device-level threats, validated through a case study on defense against False Base Station (FBS) attacks in 5G/6G systems. The authors compare various consensus strategies, demonstrate improved response efficiency and defense performance, and show that the approach can converge toward optimal power allocation under attack while maintaining auditability and robustness. The work lays groundwork for future enhancements, including dedicated consensus mechanisms and multi-modal integration, to broaden the applicability of trustworthy, LLM-supported network optimization.

Abstract

Large Language Models (LLMs) demonstrate strong potential across a variety of tasks in communications and networking due to their advanced reasoning capabilities. However, because different LLMs have different model structures and are trained using distinct corpora and methods, they may offer varying optimization strategies for the same network issues. Moreover, the limitations of an individual LLM's training data, aggravated by the potential maliciousness of its hosting device, can result in responses with low confidence or even bias. To address these challenges, we propose a blockchain-enabled collaborative framework that connects multiple LLMs into a Trustworthy Multi-LLM Network (MultiLLMN). This architecture enables the cooperative evaluation and selection of the most reliable and high-quality responses to complex network optimization problems. Specifically, we begin by reviewing related work and highlighting the limitations of existing LLMs in collaboration and trust, emphasizing the need for trustworthiness in LLM-based systems. We then introduce the workflow and design of the proposed Trustworthy MultiLLMN framework. Given the severity of False Base Station (FBS) attacks in B5G and 6G communication systems and the difficulty of addressing such threats through traditional modeling techniques, we present FBS defense as a case study to empirically validate the effectiveness of our approach. Finally, we outline promising future research directions in this emerging area.

Figures (6)

• Figure 1: Compared with a single LLM and MultiLLM, Trustworthy MultiLLMN has distributed characteristics and is less likely to be disturbed by malicious LLMs. It has higher network stability and robustness, and can provide trusted services for user.
• Figure 2: The blockchain-driven Trustworthy MultiLLMN. The response provided by a certain LLM to a user will be verified and compared by all LLMs in Trustworthy MultiLLMN. Thus, it ensures that the blockchain-driven network can provide the user with the highest quality and trusted answer.
• Figure 3: Different wireless transmission power distribution methods for LBSs. Both optimization and DL-based methods suffer from poor flexibility and portability, while the LLM-based method only needs appropriate prompts to generate power allocation strategies.
• Figure 4: Trustworthy MultiLLMN-enabled defense mechanism for the FBS attack. It can provide an LBSs power allocation method for wireless communication system to resist this attack with just a few prompts.
• Figure 5: Response time. It compares the time required for different consensus to drive the work of Trustworthy MultiLLMN.