Behavior-Based Authentication System
thesisposted on 28.04.2018, 00:00 by Taylor Means, Jared Frank
All current forms of authentication are exploitable via social engineering, theft, hacking, or replication. Due to this, a new form of authentication should be explored: behavioral. A solution to this problem would result in more secure digital environment, including physical access to computers as well as software access. The maze-solving approach presented by this project allows for multiple variables to be observed within a user, presenting many facets of behavior that can be analyzed. In order to solve this problem, enough parameters must be collected and contrasted against one another in order to tell different humans apart from each other based on how they solve a maze. Other methods of currently existing authentication rely on what you own (physical keys), what you know (passwords), and what you have (biometrics). By creating a randomly generated maze and having an observer AI object keep track of how different users solve a maze, we are able to tell two different users apart from one another to a similar degree of accuracy as other methods do. Our AI factors in variables such as time spent moving the player, time spent not moving, backtracking, strategy, and more.