What if your smartphone could recognize you by how you touch the screen?

Traditional authentication systems, such as PINs and passwords, are vulnerable to various attacks. A new study proposes a robust framework using machine learning to recognize touch behavior on smartphones. By analyzing trajectories, pressure, and touch speed, the system identifies unique patterns for each user. Using algorithms like Random Forest, Gradient Boosting, and K-Nearest Neighbors, accuracy reaches 95.2%. Additionally, the system is resilient to various attacks, maintaining high precision even in variable conditions. This method promises to enhance smartphone security in a user-friendly and scalable way.

The proposed framework combines touch pattern recognition with advanced machine learning techniques. TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) is used to select the most relevant features, such as trajectory curvature, touch pressure, and speed. These data are then processed by three machine learning algorithms: Random Forest, Gradient Boosting Machines, and K-Nearest Neighbors. The Random Forest model has proven to be the most effective, with an F1-Score of 95.1% and an accuracy of 95.2%. To further enhance the system's resilience, data perturbation techniques, such as temporal scaling and noise insertion, have been applied. This approach ensures that the system is resilient to various attacks, including those based on environmental variations and spoofing, with a detection rate of 96%.