Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble
Smartphone model detection through sensor data is important for enhancing security protocols, preventing device fraud, and ensuring authorized service access. While extensive research has utilized sensors like cameras, microphones, accelerometers, and magnetometers for device fingerprinting, gyrosco...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10811928/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841546107072544768 |
|---|---|
| author | Erkan Kiymik Ali Emre Ozturk |
| author_facet | Erkan Kiymik Ali Emre Ozturk |
| author_sort | Erkan Kiymik |
| collection | DOAJ |
| description | Smartphone model detection through sensor data is important for enhancing security protocols, preventing device fraud, and ensuring authorized service access. While extensive research has utilized sensors like cameras, microphones, accelerometers, and magnetometers for device fingerprinting, gyroscope data has remained largely unexplored for model detection due to its high susceptibility to noise from small vibrations and mechanical imperfections. This study investigates the use of gyroscope data alone for smartphone model detection. Leveraging the Google Smartphone Decimeter Challenge 2023–2024 dataset, which provides real-world gyroscope data from multiple smartphones mounted identically on vehicles during various driving tests, the challenging task of distinguishing between smartphone models under nearly identical motion conditions is addressed. A fine-tuned WaveNet model is employed to analyze the sequential nature of the gyroscope data, and an EfficientNetV2 model captures intricate time-frequency representations using Continuous Wavelet Transform (CWT) with the Morlet wavelet. Combining these models in an ensemble framework enhanced with an attention mechanism gives an accuracy of 99.01% using just 1–2 seconds of gyroscope data. This performance suggests that gyroscope data alone can be effective for model identification, even under challenging real-world conditions. These findings indicate the potential of gyroscope data in device fingerprinting and may provide a foundation for future advancements in mobile device security and authentication. |
| format | Article |
| id | doaj-art-0bad63e1ed1b4770b254388941a9c0f7 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-0bad63e1ed1b4770b254388941a9c0f72025-01-11T00:00:58ZengIEEEIEEE Access2169-35362024-01-011219548319550410.1109/ACCESS.2024.352122610811928Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 EnsembleErkan Kiymik0https://orcid.org/0000-0002-6383-1878Ali Emre Ozturk1https://orcid.org/0000-0001-5904-9931Department of Electrical and Electronics Engineering, Hasan Kalyoncu University, Gaziantep, TürkiyeDepartment of Electrical and Electronics Engineering, Hasan Kalyoncu University, Gaziantep, TürkiyeSmartphone model detection through sensor data is important for enhancing security protocols, preventing device fraud, and ensuring authorized service access. While extensive research has utilized sensors like cameras, microphones, accelerometers, and magnetometers for device fingerprinting, gyroscope data has remained largely unexplored for model detection due to its high susceptibility to noise from small vibrations and mechanical imperfections. This study investigates the use of gyroscope data alone for smartphone model detection. Leveraging the Google Smartphone Decimeter Challenge 2023–2024 dataset, which provides real-world gyroscope data from multiple smartphones mounted identically on vehicles during various driving tests, the challenging task of distinguishing between smartphone models under nearly identical motion conditions is addressed. A fine-tuned WaveNet model is employed to analyze the sequential nature of the gyroscope data, and an EfficientNetV2 model captures intricate time-frequency representations using Continuous Wavelet Transform (CWT) with the Morlet wavelet. Combining these models in an ensemble framework enhanced with an attention mechanism gives an accuracy of 99.01% using just 1–2 seconds of gyroscope data. This performance suggests that gyroscope data alone can be effective for model identification, even under challenging real-world conditions. These findings indicate the potential of gyroscope data in device fingerprinting and may provide a foundation for future advancements in mobile device security and authentication.https://ieeexplore.ieee.org/document/10811928/Gyroscopeaccelerometerfingerprintsmartphoneneural networkmachine learning |
| spellingShingle | Erkan Kiymik Ali Emre Ozturk Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble IEEE Access Gyroscope accelerometer fingerprint smartphone neural network machine learning |
| title | Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble |
| title_full | Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble |
| title_fullStr | Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble |
| title_full_unstemmed | Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble |
| title_short | Gyroscope-Based Smartphone Model Identification via WaveNet and EfficientNetV2 Ensemble |
| title_sort | gyroscope based smartphone model identification via wavenet and efficientnetv2 ensemble |
| topic | Gyroscope accelerometer fingerprint smartphone neural network machine learning |
| url | https://ieeexplore.ieee.org/document/10811928/ |
| work_keys_str_mv | AT erkankiymik gyroscopebasedsmartphonemodelidentificationviawavenetandefficientnetv2ensemble AT aliemreozturk gyroscopebasedsmartphonemodelidentificationviawavenetandefficientnetv2ensemble |