Human Trait Analysis via Machine Learning Techniques for User Authentication

Machine learning is an extremely important technique that has become heavily used in different types of applications such as detection systems for fraud, intrusion or fault and monitoring systems for health or computer. Human trait analysis and identification is a field of research that needs a strong implementation for machine learning. Human trait analysis provides a tool with which human identification factors can be verified. Currently, detail aspects of human behavior are digitally and continuously logged in Big Data based platforms such as Twitter and Facebook. This continuous flow of high-volume data requires sophisticated data analysis to examine huge amounts of behavioral evidence so that human traits can be modeled. This paper proposes an innovative technique for human trait analysis that fits the needs for user's identity verification. The pioneering work of this technique is in the distinction of the normal and abnormal actions of the users, where the focus is given to these abnormal actions to establish security potential profiles. The data analysis and prediction of the proposed technique is based on the concept of machine learning and uses several models based on four techniques; K-means, Hidden Markov Model (HMM), Auto-Encoder Neural Network, and Gaussian Distribution. Experiments have been carried out in four main phases: prediction of rare user actions, filter security potential actions, build/update a user profile, and generate a real-time (i.e. just in time) set of challenging questions. Real-world scenarios are considered to demonstrate the benefits of these challenging questions in building secure knowledge-based user authentication systems.