Automatically Gender Identification from Smartphone with Undefined Keystroke Sequence

User identification is considered an important part of any online platform. The findings of this study shed light on determining user gender from anonymous data simply from the typing pattern on the touchscreen of smartphones and tablets by filling out a questionnaire. Unlike previous analyses of structured text, the present study examines keystroke data from free, natural typing, which captures more temporal variation besides the behavior. A shift from structured to unstructured input brings forth different challenges and opportunities, which is a major contribution of the study. The influence of correct gender prediction does not consider user personalization alone, but also includes spheres such as virtual assistants, targeted advertising, and recommendations system. The researchers developed a web-based application and collected keystroke data through this system. A total of ten well-known machine learning models were used to map the typing patterns obtained from 157 applicants. The identification of such aspects of the typing pattern (time interval between key sequence to input characters) of the undefined text is remarkable. Among the models tested, the Extra Trees Classifier was observed to have the highest accuracy at 81%, beating other statistical approaches, including the Random Forest Classifier (77%), K-Nearest Neighbors (76%), and Extreme Gradient Boosting (76%), with an ability to determine the gender of unknown users based on keystrokes. As a result, this is a straightforward and useful way to recognize users in a keystroke detection system for the smartphone platform.