Features Extraction Scheme for Behavioural Biometric Authentication in Touchscreen Mobile Devices

Today, mobile devices are being widely used in personal and professional life. By increasing the popularity of touchscreen platform as an input method in mobiles phones, touch gesture behaviour is becoming more significantly important in interaction with the phone. Due to increasing demand for safer access in touchscreen mobile phones, old strategies like pins, tokens, or passwords have failed to stay abreast of the challenges. By utilizing touch gesture behaviour biometric techniques, the authentication mechanism will improve and would make it more difficult for a shoulder surfer to replay the password, even if he observes the entire gesture. The purpose of this research is to extract different touch gesture behavioural features and find the best feature using a discriminative classifier to improve the current touch-gesture authentication techniques. The results of this research show that forward subset features selection scheme with Random Forest classification technique outperform the existing touch-gesture authentication methods in terms of false accept rate, equal error rate and authentication accuracy by 1.5%, 4.65%, and 1.67% respectively.