People relying on personal devices for personal information is the new normal. As the ecosystem of such devices is more connected, they have developed and become highly dynamic and convoluted. Current user authentication solutions in action like Password and PIN might not be reliable enough for user data security. This paper examines how user behavioural recognition can mitigate user authentication challenges. We have used two open-source datasets of user behavioural biometrics (touch dynamics and periocular traits). Later, fusion was performed at the feature level and incorporated distance measure techniques (Euclidean, Manhattan, Cosine, Minkowski) to determine whether the user was genuine or an imposter. We have also performed multi-data processing methods such as Principal Component Analysis (PCA) & Linear Discriminant Analysis (LDA). Our proposed model achieved recognition accuracy of 94.30% using the median statistical index, Manhattan distance measure, LDA feature reduction, and KNN-supervised machine learning technique.