SENTINEL-DL: a forensic framework for device attribution using motion sensor data.

Area of Science:

• Digital Forensics
• Sensor Data Analysis
• Machine Learning

Background:

• Digital evidence analysis often relies on data stored on devices.
• Associating motion-based evidence with specific device models presents a challenge.
• Existing forensic methods may lack robust device fingerprinting capabilities.

Purpose of the Study:

• To introduce SENTINEL-DL, a novel forensic framework for device association using motion sensor data.
• To leverage intrinsic motion signatures from accelerometers for forensic analysis.
• To enhance digital fingerprinting and forensic authentication for motion evidence.

Main Methods:

• Utilized built-in 3D accelerometers to profile tamper-resistant intrinsic motion signatures.
• Employed machine learning models, including random forest (RF), deep neural networks (DNN), and convolutional neural networks (CNN).
• Matched unknown sensory data against a reference database of device profiles.

Main Results:

• Achieved a true positive rate (TPR) of 93.99% for smartphones and 92.65% for smartwatches.
• Reported a false acceptance rate (FAR) of 0.66% for smartphones and 1.22% for smartwatches.
• Demonstrated overall accuracy of 98.76% for smartphones and 98.97% for smartwatches.

Conclusions:

• SENTINEL-DL provides a lightweight and dependable solution for analyzing motion sensor evidence.
• The framework offers effective digital fingerprinting and forensic authentication support.
• Motion sensor data can significantly improve device fingerprinting and forensic verification methodologies.