ChatIoT: Large Language Model-based Security Assistant for Internet of Things with Retrieval-Augmented Generation

• DOI: 10.48550/arxiv.2502.09896
• Publication Date: 2025-02-13
• Authors: Dong Ye; Yan Lin Aung; Sudipta Chattopad...; Jianying Zhou
• Keywords: FOS: Computer and information sciences; Computer Science - Cryptography and Security; Cryptography and Security (cs.CR)

Internet of Things (IoT) has gained widespread popularity, revolutionizing industries and daily life. However, it also emerged as a prime target for attacks. Numerous efforts have been made to improve IoT security, substantial security threat information, such datasets reports, developed. existing research often falls short in leveraging these insights assist or guide users harnessing practices clear actionable way. In this paper, we propose ChatIoT, large language model (LLM)-based assistant designed disseminate intelligence. By the versatile property retrieval-augmented generation (RAG), ChatIoT successfully integrates advanced understanding reasoning capabilities LLM with fast-evolving information. Moreover, develop an end-to-end data processing toolkit handle heterogeneous datasets. This converts various formats into retrievable documents optimizes chunking strategies efficient retrieval. Additionally, define set common use case specifications generating answers aligned users' specific needs expertise levels. Finally, implement prototype conduct extensive experiments different LLMs, LLaMA3, LLaMA3.1, GPT-4o. Experimental evaluations demonstrate that can generate more reliable, relevant, technical in-depth most cases. When evaluating LLaMA3:70B, improves above metrics by over 10% on average, particularly relevance technicality, compared using LLMs alone.