Identification of Mobile Phones Using the Built-In Magnetometers Stimulated by Motion Patterns.

Area of Science:

• Electrical Engineering
• Computer Science
• Cybersecurity

Background:

• Mobile phones contain built-in magnetometers, electronic components widely deployed in mass-market devices.
• Physical variations in these magnetometers result in unique digital outputs, enabling device identification.
• Prior research has explored identifying phones via cameras, microphones, and RF components.

Purpose of the Study:

• To investigate the feasibility of uniquely identifying mobile phones using their internal magnetometers.
• To develop an inexpensive method for stimulating and analyzing magnetometer output for device fingerprinting.

Main Methods:

• An experimental setup was designed with a rotating platform for the mobile phone and a fixed magnet.
• Magnetometer data was recorded as the phone passed the magnet, with experiments repeated over six days for consistency.
• Statistical features were extracted from synchronized and correlated magnetometer outputs to create device fingerprints.
• A Support Vector Machine (SVM) machine learning algorithm was employed for classification.

Main Results:

• Accurate classification was achieved between different mobile phone models and brands (inter-model).
• Classification accuracy for individual phones of the same model (intra-model) was only slightly better than random chance.
• The findings highlight the potential and limitations of magnetometer-based mobile phone identification.

Conclusions:

• Mobile phone identification via built-in magnetometers is feasible, particularly for differentiating between models.
• Intra-model identification presents significant challenges due to subtle variations within the same model.
• Further research may be needed to improve the accuracy of identifying individual devices of the same model.