Self-service password reset

Self-service password reset (SSPR) is defined as any process or technology that allows users who have either forgotten their password or triggered an intruder lockout to authenticate with an alternate factor, and repair their own problem, without calling the help desk. It is a common feature in identity management software and often bundled in the same software package as a password synchronization capability.

Typically users who have forgotten their password launch a self-service application from an extension to their workstation login prompt, using their own or another user's web browser, or through a telephone call. Users establish their identity, without using their forgotten or disabled password, by answering a series of personal questions, using a hardware authentication token, responding to a notification e-mail or, less often, by providing a biometric sample such as voice recognition. Users can then either specify a new, unlocked password, or ask that a randomly generated one be provided.

Self-service password reset expedites problem resolution for users "after the fact", and thus reduces help desk call volume. It can also be used to ensure that password problems are only resolved after adequate user authentication, eliminating an important weakness of many help desks: social engineering attacks, where an intruder calls the help desk, pretends to be the intended victim user, claims to have forgotten the account password, and asks for a new password.

Rather than merely asking users to answer security questions, modern password reset systems may also leverage a sequence of authentication steps:

Despite the benefits, a self-service password reset that relies solely on answers to personal questions can introduce new vulnerabilities, since the answers to such questions can often be obtained by social engineering, phishing techniques or simple research. While users are frequently reminded never to reveal their password, they are less likely to treat as sensitive the answers to many commonly used security questions, such as pet names, place of birth or favorite movie. Much of this information may be publicly available on some users' personal home pages. Other answers can be elicited by someone pretending to conduct an opinion survey or offering a free dating service. Since many organizations have standard ways of determining login names from real names, an attacker who knows the names of several employees at such an organization can choose one whose security answers are most readily obtained.

This vulnerability is not strictly due to self-service password reset—it often exists in the help desk prior to deployment of automation. Self-service password reset technology is often used to reduce this type of vulnerability, by introducing stronger caller authentication factors than the human-operated help desk had been using prior to deployment of automation.

In September 2008, the Yahoo e-mail account of Governor of Alaska and Vice President of the United States nominee Sarah Palin was accessed without authorization by someone who was able to research answers to two of her security questions, her zip code and date of birth and was able to guess the third, where she met her husband. This incident clearly highlighted that the choice of security questions is very important to prevent social engineering attacks on password systems.

Jakobsson, Stolterman, Wetzel, and Yang proposed to use preferences to authenticate users for password reset. The underlying insights are that preferences are stable over a long period of time, and are not publicly recorded. Their approach includes two phases---setup and authentication. During the setup, a user is asked to select items that they either like or dislike from several categories of items which are dynamically selected from a big candidate set and are presented to the user in a random order. During the authentication phase, users are asked to classify their preferences (like or dislike) for the selected items displayed to them in a random order. Jakobsson, Stolterman, Wetzel, and Yang evaluated the security of their approach by user experiments, user emulations, and attacker simulations.