Abstract
This paper introduces SENTINEL-DL-a novel forensic framework which leverages accelerometer sensory data to associate motion-based digital evidence to its corresponding smartphone or smartwatch models. SENTINEL-DL analyzes robust tamper-resistant intrinsic motion signatures (profiled using built-in 3D accelerometers) to establish device associations. Technically speaking, it leverages small differences in linear acceleration to identify and associate the readings with its generating device. SENTINEL-DL utilizes machine learning models including random forest (RF), deep neural networks (DNN) and convolutional neural networks (CNN) to drive its association during the matching process, i.e., unknown sensory data against a reference database containing device profiles from known sources. The results of empirical tests show that SENTINEL-DL for smartphones and smartwatches, respectively, achieves a true positive rate (TPR) of 93.99% and 92.65%, a false acceptance rate (FAR) of 0.66% and 1.22%, and an overall accuracy of 98.76% and 98.97%. SENTINEL-DL being light-weight promises investigators a dependable analysis solution for motion sensor evidence while providing digital fingerprinting capabilities and forensic authentication support. The research demonstrates how motion sensor data can be utilized in digital forensic investigations to develop improved device fingerprinting and forensic verification methodologies.