Smartphone user authentication is still an open challenge because the balance between both security and usability is indispensable. To balance between them, active authentication is one way to overcome the problem. In this paper, we tackle to improve the accuracy of active authentication by adopting online learning with touch pressure. In recent years, it becomes easy to use the smartphones equipped with pressure sensor so that we have confirmed the effectiveness of adopting the touch pressure as one of the features to authenticate. Our experiments adopting online AROW algorithm with touch pressure show that equal error rate (EER), where the miss rate and false rate are equal, is reduced up to one-fifth by adding touch pressure feature. Moreover, we have confirmed that training with the data from both sitting posture and prone posture archives the best when testing variety of postures including sitting, standing and prone, which achieves EER up to 0.14%.