Towards Robust Mobile Authentication: Behavioral Biometrics and Optimized Classification for Continuous User Recognition
Srinivasa Rao Bittla, Sunil Kumar, Srimaan Yarram · 2025
The utilisation of soft keyboard typing behaviour as a behavioural biometric for continuous user recognition is the main contribution of this study. This is accomplished by analysing data acquired from the touchscreen and motion sensors to determine the user's typing style and grip style on the phone. The processing speed and precision of recognition provide additional challenges to mobile device authentication. An optimised and speedy data classification system based on a class attention layer (CAL) in CapsNet is proposed. Further to the study's contributions, this structure uses the Slime Mould Algorithm (SMA) to choose features, which provides fast and very accurate classification results. Research shows that a classification accuracy of up to 98% is achievable when identifying users. Superior security is provided by continuous authentication systems as contrasted with one-time authentication systems. However, the results produced by these methods may not always be spot-on. The creation of an effective software architecture is crucial for resolving this issue. Along these lines, the paper adds a description of how to build a incessant authentication scheme utilising the architecture that was built.