Leveraging an Active Directory for the Generation of Honeywords

Honeywords, fake passwords that when used by an adversary are set to trigger an alarm, is one way of detecting security breaches. For them to be effective, however, they must resemble real passwords as closely as possible and thus, the construction of the honeywords is crucial. In this thesis, a new model for generating honeywords, PII-Syntax, is presented that was built in part on a previous model but reworked and adapted to meet new requirements. The purpose of the study was to investigate whether an Active Directory, (AD) could be used as a resource in the construction of honeywords. The assumption was that the AD contains information about real system users that could be leveraged to create high-quality honeywords because of the very fact that they are based on actual users. It is a well-known fact that many users have a natural inclination towards incorporating personal information when choosing their passwords, information that can be leveraged by an adversary making the passwords easier to retrieve. The proposed model capitalizes on this fact and bases the honeyword generation process on users’ personally identifiable information, PII. The motivation for this is to enhance the quality of the honeywords, i.e. making them more plausible from the perspective of the adversary. The resulting model performed equally well or better than all existing honeyword generation algorithms to which it was compared with regard to flatness, DoS resistivity, multiple system vulnerability and storage cost. The most important contribution, however, is the inclusion of users’ personal information in the generation of the honeywords that ultimately help strengthen the security of password-based authentication systems. Contributions from this thesis include a novel manner in which to approach a well-known problem, both in a theoretical as well as a practical sense: PII-Syntax is a new honeyword generation algorithm that apart from performing equally well or better than previous algorithms brings an added value of believability to the generated honeywords because of the inclusion of users’ personal information found in an AD