Privacy Leakage: Gender Identification Based on Built In Smartphone Accelerator Usage

Personal information is a kind of sensitive data that users prefer keeping in private. However, with the development of Mobile Internet, various novel applications generate a large amount of data, which could be used to correlate the users and further infer their private information. This paper argues that smartphone application usage would incur potential privacy issue, and proposes a scheme to verify a case study that gender information leaks while the built-in smartphone accelerator is used. In particular, an Android OS application was firstly developed to collect the data generated by smartphone accelerator. Then, by extracting multiple features from the collected acceleration data, constructing an adjacent matrix to correlate users and features, and leveraging the Support Vector Machine (SVM) algorithm, users are classified into either males or females. Finally, while evaluating the scheme with real world data, experimental results show that, just by leveraging a four-second accelerator data segment of each user, the accuracy and F1 value can both achieve 84\%, respectively. This paper reveals the potential privacy leakage in smartphone usage, and provides valuable input for smartphone users and security service firms seeking to improve their products.