Biometric Fraud Detection System

As the internet and technology become integral to daily life, user authentication remains a critical challenge, especially with the rise of social engineering. Biometrics, particularly physiological and behavioral, offer promising solutions. Keystroke dynamics, or typing biometrics, identifies users based on their typing patterns. This research evaluates the effectiveness of keystroke dynamics by analyzing existing literature and proposes a fraud detection system combining both physiological and behavioral biometrics. We tested a Vanilla Neural Network (Vanilla-NN) that improves performance in fixed-text keystroke dynamics and propose a novel Long Short-Term Memory (LSTM) attention model for free-text dynamics, delivering promising results.