TSCA: Enhancing Smartphone Security with a Touchpoints-Sensitive and Context-Aware Model for Touchscreen-Based Authentication

With the widespread use of smartphones, more and more private information is stored on them, which requires a more secure authentication method. Continuous authentication methods show potential for privacy protection due to their continuous and transparent nature, as opposed to the entry-point authentication methods like PIN codes and facial recognition. Users' touchscreen behavior, being both common and unique, has been widely studied in the field. However, previous methods based on manual features and deep learning have certain shortcomings when mining patterns from touchscreen behaviors that are beneficial for authentication. First, the operating habits of different users lead to different importance of touch points in a single touchscreen swipe. Second, extracting features from each touchscreen swipe independently ignores useful information from the context. To address this problem, we propose a novel Touchpoints-Sensitive and Context-Aware model called TSCA, which is able to adaptively assign weights to different touch points in a touchscreen swipe while also taking into account information from the context of the touchscreen swipe. We evaluate the TSCA model on two publicly available datasets and show that the TSCA is able to deliver significant performance gains compared to baseline methods while reducing model complexity.