BehavePassDB: Public Database for Mobile Behavioral Biometrics and Benchmark Evaluation

• We present a new HCI database, BehavePassDB, with novel data collection approach. exploit the touchscreen and background sensor for mobile authentication. employ Deep Learning approach based on LSTM to benchmark BehavePassDB. evaluate two different impostor scenarios considering real-life use cases. influence of device bias authentication performance. Mobile behavioral biometrics have become popular topic research, reaching promising results in terms authentication, exploiting multimodal combination data. However, there is no way knowing whether state-of-the-art classifiers literature can distinguish between notion user device. In this article, we structured into separate acquisition sessions tasks mimic most common aspects Human-Computer Interaction (HCI). BehavePassDB acquired through dedicated app installed subjects devices, also including case users same evaluation. propose standard experimental protocol research community perform fair comparison approaches state art 1 https://github.com/BiDAlab/MobileB2C_BehavePassDB/ . system Long-Short Term Memory (LSTM) architecture triplet loss modality fusion at score level.