How Do Users Chain Email Accounts Together? [IFIP SEC 2021]

   Authors: Lydia Kraus and Mária Švidroňová and Elizabeth Stobert

 Primary contact: Lydia Kraus <lydia.kraus@mail.muni.cz>

 Conference: IFIP SEC 2021

   DOI: TBD

Pre-print PDF   Artifacts   BiBTeX
@InProceedings{2021-ifipsec-kraus,
Title         = {How Do Users Chain Email Accounts Together?},
Author        = {Lydia Kraus and Maria Svidronova and Elizabeth Stobert},
BookTitle     = {Accepted at IFIP International Conference on ICT Systems Security and Privacy Protection},
Pages         = {},
Publisher     = {Springer, Cham},
Year          = {2021},
Keywords      = {usablesec},
crocsweb      = {https://crocs.fi.muni.cz/papers/ifipsec2021},

}

Abstract

Recovery connections between email accounts can be exploited in manual hijacking attacks as has been shown in several incidents during the last years. Yet little is known about users' practices of chaining email accounts together. We conducted a qualitative interview study with 23 students in which they shared their email recovery and forwarding settings with us. Altogether, we collected and analyzed information about 138 different email accounts. We used this data to map email account topologies and analyzed these topologies for recurring patterns. We found that users often make poor configuration decisions in their email recovery setups, and often create patterns in their email recovery topologies that result in security vulnerabilities. Patterns such as loops (seen in more than a quarter of our topologies) could be easily exploited in a targeted attack. We conclude that users need better guidance about how to use email based recovery settings in a robust way.

  • Our results show that email account topologies are diverse, but that many include elements of line and loop topologies.
  • Loop topologies are especially concerning, as they allow attackers who already have access to one account to easily gain access to a further account.
  • We also found that users created other vulnerabilities in their topologies by placing the final recovery nodes in inaccessible accounts, or by using accounts owned by somebody else as a recovery option.
  • Users also tended to keep accounts with physical recovery options (which could be strong recovery options) separate from the rest of the recovery topology.
  • Participants were often unaware of the connections between their accounts and only realized during the study that there are recovery links they are not happy with.