Exploiting Contactless Side Channels in Wireless Charging Power Banks for User Privacy Inference via Few-shot Learning
Open Access
- 2 October 2023
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 1-15
Abstract
Recently, power banks for smartphones have begun to support wireless charging. Although these wireless charging power banks appear to be immune to most reported vulnerabilities in either power banks or wireless charging, we have found a new contactless wireless charging side channel in these power banks that leaks user privacy from their wireless charging smartphones without compromising either power banks or victim smartphones. We have proposed BankSnoop to demonstrate the practicality of the newly discovered wireless charging side channel in power banks. Specifically, it leverages the coil whine and magnetic field disturbance emitted by a power bank when wirelessly charging a smartphone and adopts the few-shot learning to recognize the app running on the smartphone and uncover keystrokes. We evaluate the effectiveness of BankSnoop using commodity wireless charging power banks and smartphones, and the results show it achieves over 90% accuracy on average in recognizing app launching and keystrokes. It also presents high adaptability when apply to different smartphone models, power banks, etc., achieving over 85% accuracy with 10-shot learning.Keywords
Funding Information
- CityU APRC grant (9610563)
- Hong Kong RGC (CityU 21219223, C1029-22G, CityU 21201420, CityU 11201422, GRF PolyU 15224121)
- the China Postdoctoral Science Foundation (2023M732791)
- NSFC (62101471)
- NSF of Shandong Province (ZR2021LZH010)
- Shenzhen Science and Technology Funding Fundamental Research Program (2021Szvup126)
This publication has 10 references indexed in Scilit:
- MagEar2022
- Eavesdropping user credentials via GPU side channels on smartphones2022
- A Simple and Fast Human Activity Recognition System Using Radio Frequency Energy Harvesting2021
- Deep Transfer Learning for Cross-domain Activity Recognition2018
- DolphinAttack2017
- Deep convolutional feature transfer across mobile activity recognition domains, sensor modalities and locations2016
- How is energy consumed in smartphone display applications?2013
- Nearest Neighbor Value InterpolationInternational Journal of Advanced Computer Science and Applications, 2012
- Chameleon2011
- Learning decision tree classifiersACM Computing Surveys, 1996