A Secure LSTM-Based Indoor Localization Method with Multi-Source Trajectory Fingerprint

Fingerprint localization maps real-time wireless signals to physical locations, addressing accuracy degradation in complex environments. However, its effectiveness heavily relies on the uniqueness and stability of the fingerprints. This paper presents FusionTrack, an innovative multi-source trajectory fin-gerprinting approach that combines 5G signals and geomagnetic field strength to enhance fingerprint quality. The method employs an LSTM-based two-stage approach: first, it estimates the device's region ID and integrates it into the fingerprint, followed by fine-grained localization. During this process, a self-correction algorithm based on feedback is applied to prevent error propagation from incorrect region ID predictions. Real-world evaluations show that the method achieves a positioning accuracy of 0.72 meters, improving by 85.2 % over the baseline, with outstanding robustness and precision. Additionally, FusionTrack incorporates advanced security mechanisms to ensure data integrity and protect location privacy, making it a reliable solution for secure and accurate indoor positioning.