Doctor of Philosophy (Ph.D.)
Password has been the dominant authentication scheme for more than 30 years, and it will not be easily replaced in the foreseeable future. However, password authentication has long been plagued by the dilemma between security and usability, mainly due to human memory limitations. For example, a user often chooses an easy-to-guess (weak) password since it is easier to remember. The ever increasing number of online accounts per user even exacerbates this problem. In this dissertation, we present four research projects that focus on the security of password authentication and its ecosystem. First, we observe that personal information plays a very important role when a user creates a password. Enlightened by this, we conduct a study on how users create their passwords using their personal information based on a leaked password dataset. We create a new metric---Coverage---to quantify the personal information in passwords. Armed with this knowledge, we develop a novel password cracker named Personal-PCFG (Probabilistic Context-Free Grammars) that leverages personal information for targeted password guessing. Experiments show that Personal-PCFG is much more efficient than the original PCFG in cracking passwords. The second project aims to ease the password management hassle for a user. Password managers are introduced so that users need only one password (master password) to access all their other passwords. However, the password manager induces a single point of failure and is potentially vulnerable to data breach. To address these issues, we propose BluePass, a decentralized password manager that features a dual-possession security that involves a master password and a mobile device. In addition, BluePass enables a hand-free user experience by retrieving passwords from the mobile device through Bluetooth communications. In the third project, we investigate an overlooked aspect in the password lifecycle, the password recovery procedure. We study the password recovery protocols in the Alexa top 500 websites, and report interesting findings on the de facto implementation. We observe that the backup email is the primary way for password recovery, and the email becomes a single point of failure. We assess the likelihood of an account recovery attack, analyze the security policy of major email providers, and propose a security enhancement protocol to help securing password recovery emails by two factor authentication. \newline Finally, we focus on a more fundamental level, user identity. Password-based authentication is just a one-time checking to ensure that a user is legitimate. However, a user's identity could be hijacked at any step. For example, an attacker can leverage a zero-day vulnerability to take over the root privilege. Thus, tracking the user behavior is essential to examine the identity legitimacy. We develop a user tracking system based on OS-level logs inside an enterprise network, and apply a variety of techniques to generate a concise and salient user profile for identity examination.
© The Author
Li, Yue, "On Enhancing Security of Password-Based Authentication" (2019). Dissertations, Theses, and Masters Projects. Paper 1563898928.