International Association
for Cryptologic Research

Password managers have gained significant popularity and are widely recommended as an effective means of enhancing user security. However, current cloud-based architectures assume that password manager providers are trusted entities. This assumption is never questioned because such password managers are operated by their own designers, which are therefore judge and jury. This exposes users to significant risks, as a malicious provider could perform covert actions without being detected to access or alter users' credentials. This exposes users to significant risks, as a malicious provider could perform covert actions without being detected to access or alter the credentials of users. Most password managers rely solely on the strength of a user-chosen master password. As a result, a covert adversary could conceivably perform large-scale offline attacks to recover credentials protected by weak master passwords. Even more concerning, some password managers do not encrypt credentials on users' devices, transmitting them in plaintext before encrypting them server-side, e.g., Google, in its default configuration. On the other hand, key-protected password managers, e.g., KeePassXC, are less commonly used, as they lack functionality for synchronizing credentials across multiple devices.

In this paper, we establish a comprehensive set of security properties that should be guaranteed by any cloud-based password manager. We demonstrate that none of the widely deployed mainstream password managers fulfill these fundamental requirements. Nevertheless, we argue that it is feasible to design a solution that is resilient against covert adversaries while allowing users to synchronize their credentials across devices. To support our claims, we propose a password manager design that fulfills all the required properties.