ChatTracer: Large Language Model Powered Real-time Bluetooth Device Tracking System

TL;DR

This paper explores embedding large language models into wireless sensor networks to provide real-time Bluetooth device tracking and natural language interaction. It introduces ChatTracer, a system with BLE sniffing nodes, a database, and a fine-tuned LLM (Mistral-7B) trained via supervised fine-tuning and RLHF. Key innovations include a BLE packet grouping algorithm, AoA/RSSI/CFO feature extraction, and device payload decoding across Android and Apple devices, integrated into an end-to-end pipeline. Experimental results across multiple realistic environments show ChatTracer outperforms state-of-the-art localization baselines and offers an intuitive interface for user queries, highlighting the potential for LLM-assisted WSN.

Abstract

Large language models (LLMs) have transformed the way we interact with cyber technologies. In this paper, we study the possibility of connecting LLM with wireless sensor networks (WSN). A successful design will not only extend LLM's knowledge landscape to the physical world but also revolutionize human interaction with WSN. To the end, we present ChatTracer, an LLM-powered real-time Bluetooth device tracking system. ChatTracer comprises three key components: an array of Bluetooth sniffing nodes, a database, and a fine-tuned LLM. ChatTracer was designed based on our experimental observation that commercial Apple/Android devices always broadcast hundreds of BLE packets per minute even in their idle status. Its novelties lie in two aspects: i) a reliable and efficient BLE packet grouping algorithm; and ii) an LLM fine-tuning strategy that combines both supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF). We have built a prototype of ChatTracer with four sniffing nodes. Experimental results show that ChatTracer not only outperforms existing localization approaches, but also provides an intelligent interface for user interaction.

Figures (3)

• Figure 1: ChatTracer's system architecture.
• Figure 2: Location error distribution of ChatTracer in comparison with model-based localization.
• Figure 3: Illustration of ChatTracer's trajectory accuracy in three scenarios.