Secure Indoor Positioning System Using Boundary Attack Based With Mobile Application

This paper presents the development of a secure Indoor Positioning System (IPS) resistant to Boundary Attacks, leveraging Channel State Information (CSI) fingerprint data. Traditional IPS technologies, while achieving high accuracy, often neglect security and privacy, making them vulnerable to various attacks. This study focuses on enhancing the security of IPS by employing the eXtreme Gradient Boosting (XGBoost) model and integrating it with a mobile application for real-time location tracking. The research methodology encompasses data collection from Tower 2 ITS using the ESP32 CSI Toolkit, data preprocessing, model training, and testing against Boundary Attacks. The results demonstrate a significant improvement in the model's resilience to attacks. The initial model accuracy of 75.92% for the Tower 2 ITS dataset dropped to 5.72% after the attack but was restored to 61.68% with defense training. Similarly, the comparison dataset model's accuracy decreased from 99.48% to 0.52% post-attack but improved to 99.77% with the implemented defenses. Additionally, the study highlights the implementation of a mobile application using Flask, which receives and displays CSI data predictions. The application showcases the practical applicability of the developed IPS model in real-world scenarios. The research findings emphasize the necessity of incorporating robust security measures in IPS development to mitigate the risks associated with data tampering and adversarial attacks.