ChatTracer: Large Language Model Powered Real-time Bluetooth Device Tracking System
Large language models (LLMs), ItagPro exemplified by OpenAI ChatGPT and ItagPro Google Bard, iTagPro support have transformed the way in which we work together with cyber technologies. On this paper, we examine the potential of connecting LLM with wireless sensor networks (WSN). A successful design is not going to only lengthen LLM’s knowledge landscape to the physical world but additionally revolutionize human interaction with WSN. To the top, affordable item tracker we present ChatTracer, an LLM-powered actual-time Bluetooth gadget tracking system. ChatTracer contains three key components: iTagPro shop an array of Bluetooth sniffing nodes, a database, and a wonderful-tuned LLM. ChatTracer was designed primarily based on our experimental commentary that commercial Apple/Android units always broadcast lots of of BLE packets per minute even in their idle status. We have now built a prototype of ChatTracer with 4 sniffing nodes. Experimental outcomes show that ChatTracer not solely outperforms current localization approaches, but in addition gives an clever interface for user interplay. The emergence of giant language models (LLM) has ushered in a transformative era, revolutionizing the way in which we interact with expertise and harnessing the ability of pure language processing.
So far, to the better of our knowledge, LLM has not yet been combined with wireless sensor networks (WSN) (Hou et al., 2023; Fan et al., ItagPro 2023; Awais et al., 2023; Liu et al., 2023; Naveed et al., 2023; Zhao et al., 2023; Hadi et al., 2023; Guo et al., 2023; Raiaan et al., 2023; Demszky et al., 2023; Thapa and Adhikari, 2023). Connecting these two worlds is interesting for 2 reasons. First, from the LLM’s perspective, connecting with WSN will extend LLM’s capabilities from generating information-based information to offering contemporary, actual-time sensory info of our bodily world. Second, from the WSN’s perspective, the use of LLM will transform the interaction between people and WSN, making the sensory data more accessible and easier to grasp for end customers. On this paper, we present the first-of-its-form study on connecting LLM with WSN, with the intention of understanding the potential of LLM in the processing of sensory data from WSN and facilitating human interaction with WSN.
Specifically, we introduce ChatTracer, an LLM-powered real-time Bluetooth system monitoring system. ChatTracer has an array of radio sniffing nodes deployed in the area of interest, which keep listening to the radio signals emitted by the Bluetooth devices in the proximity. ChatTracer processes its received Bluetooth packets to extract their bodily and payload features using area information. The extracted per-packet options are saved in a database and fed into an LLM (Mistral-7B (Jiang et al., 2023)) to generate the human-like textual response to the queries from users. Our measurements show that, even within the powered-off standing, the iPhone 15 Pro Max still broadcasts about 50 BLE packets per minute. We discovered: (i) all Android devices broadcast not less than a hundred and twenty BLE packets per minute. By decoding their BLE packets, we are able to obtain their vendor info. In comparison with Android devices, Apple units transmit BLE packets more aggressively at a higher energy. Most Apple gadgets transmit 300-1500 packets per minute.
Additionally, most Apple devices have unique codes (Apple continuity) of their BLE packets, making it possible for ChatTracer to obtain their status and activity data. These findings verify the feasibility of utilizing ambient Bluetooth alerts for human tracking, iTagPro locator and lay the foundation for ChatTracer. To design and implement ChatTracer, we face two challenges. The first problem lies in grouping the information packets from individual Bluetooth gadgets. ChatTracer’s radio sniffing nodes will constantly obtain the data packets from all Bluetooth devices in the area of curiosity. One Bluetooth device may use completely different promoting addresses to ship their BLE packets and randomize their advertising addresses over time (e.g., each quarter-hour). It is crucial for ChatTracer to group the information packets from the identical Bluetooth device. Doing so will not only permit ChatTracer to infer the total number of Bluetooth devices, but it surely may even improve localization accuracy by increasing the number of BLE packets for device location inference.