AirTagged: A Dataset and Processing Framework for Heterogeneous High-Density IoT Environments

Personal Bluetooth Low Energy (BLE) trackers such as AirTag help locate lost items but can be misused for stalking. Research indicates that BLE trackers can be identified through their transmitted packets, thus offering potential for machine learning (ML) solutions. However, current packet datasets lack the scale and diversity needed for real-world applicability. This paper presents an open-source 200-hour BLE advertisement packet dataset focused on personal tags, enabling future ML-based device detection approaches. Additionally, introduces the first large-scale BLE data preprocessing framework for efficient and modular BLE packet preprocessing. The Framework is showcased on the dataset, demonstrating feature extraction, labelling, and dynamic plotting. This paper lays the groundwork for IoT device detection in high-density, heterogeneous environments, enabling future advances in BLE device classification.