Personalized Rhythm Click Based Authentication System Improvement Using a Statistical Classifier
The authors recently proposed a personalized rhythm click-based authentication system implemented using a neural network classifier. Unfortunately, the neural network classifier requires impostor patterns as training samples to train the network. It is impractical to collect impostor patterns in the real world. This paper presents a statistical classifier that solves these problems. The proposed system does not require impostor patterns to build the classifier and the computation is efficient. With the same benchmark dataset, FAR=2.46% and FRR=29.2%, in Chang et al.'s system is reduced to FAR=0.00% and FRR=0.06% in the proposed system.