Exploitable Flaws Found in Cloud-Based Password Managers

Claims by leading stand-alone password managers that their implementation of “zero knowledge encryption” means stored passwords can withstand the worst of hacker assaults are vastly overblown, say academic security researchers.

See Also: Cracking the Code: Securing Machine Identities

A team of four hailing from Switzerland’s ETH Zurich and the USI Università della Svizzera italiana found that security guarantees by cloud-based password manager software offered by Bitwarden, Dashlane and LastPass aren’t as advertised.

The three password managers – selected as representative samples, based on market share, their ability to access unobfuscated source code for the products, as well as “the richness of the offered feature set and the diversity of approaches” – advertise that password vaults should be safe even if hackers compromise the server that stores them.

All three of the vendors promise zero knowledge encryption, meaning they can’t see into the password vault. The researchers said there’s no industry-accepted definition for what this means: It’s a marketing term. Even so, “the promise is that even if someone is able to access the server, this does not pose a security risk to customers because the data is encrypted and therefore unreadable,” said Matilda Backendal, an assistant professor at USI Università della Svizzera italiana who was part of the research team.

“We have now shown that this is not the case,” she said.

Security experts have long recommended password managers, not least because they can be used to generate and manage unique passwords for every different site or service. The security challenge is that password managers can be filled with desirable information.

“Due to the large amount of sensitive data they contain, password managers are likely targets for experienced hackers who are capable of penetrating the servers and launching attacks from there,” said Kenneth Paterson, professor of computer science at ETH Zurich, who co-authored the report.

All three password managers fell short in some way. “The attacks allow us to downgrade security guarantees, violate security expectations and even fully compromise users’ accounts,” the researchers wrote in a paper set to be presented in August at the 35th annual USENIX Security Symposium being held in Baltimore.

“Worryingly, the majority of the attacks allow recovery of passwords – the very thing that the password managers are meant to protect.”

Researchers subjected each password manager to a “malicious server threat model,” through which they identified 12 different types of attacks that succeeded against Bitwarden, seven against LastPass and six against Dashlane.

Malicious servers can pose a risk due to how cloud-based password managers store and grant access to a user’s encrypted password vault in the cloud.

When a user of such a service wants to retrieve a password, they authenticate to the service provider, pull a copy of the vault onto their client and decrypt it using their master password.

Compared to a client-only password manager, cloud services may offer a variety of additional features. These can include the ability to share passwords with family or coworkers inside an organization, to access passwords in a variety of ways including through a web browser and mobile device, key recovery services and for organizations to self-host their own password servers.

The researchers said that in many cases, the test server they created to mimic malicious behavior not only gave them access to a user’s stored passwords, but let them change the stored password.

“We were surprised by the severity of the security vulnerabilities,” said Paterson.

They grouped the flaws they found into four categories: exploits of key escrow features that support single sign-in logins and account recovery, exploits of vault integrity, exploits of sharing features and exploiting backwards compatibility features.

The research team notified all of the vendors about their findings and shared proof-of-concept exploits, set a 90 day public disclosure deadline and said they worked closely to help as they prepared fixes. After notifying Bitwarden in January 2025, LastPass in June 2025 and Dashlane in August 2025, the researchers subsequently pushed back the disclosure deadline after a request from LastPass, which last July also awarded them two bug bounties and continued to keep them closely apprised of its efforts to patch the flaws.

“For the most part, the providers were cooperative and appreciative, but not all were as quick when it came to fixing the security vulnerabilities,” Paterson said.

Information Security Media Group approached Bitwarden, Dashlane and LastPass for comment. The researchers said all of the vendors planned to address the vulnerabilities and their response publicly.

“Dashlane found no evidence of exploitation related to these issues,” the company said and pointed to a blog post it published on Monday responding to the findings and detailing its fix.

“It’s also important to note that the attacks identified by the researchers require full compromise of a password manager’s servers, paired with a highly sophisticated threat actor able to execute cryptographic attacks, and for certain findings, either specific circumstances and/or an extremely significant window of time,” it said.

The researchers also conducted an initial assessment of 1Password, and said they found – and reported – ways its software could be exploited using a malicious server. They also noted that the software does require a “secret key” to access a vault – not just a master password – which should safeguard it against brute-force attack risks found in the other password managers. 1Password said it’s reviewed the research and found it doesn’t detail any attacks not already documented in the company’s own security design white paper, which notes: “At present, there’s no robust method for a user to verify the public key they’re encrypting data to belongs to their intended recipient. As a consequence, it would be possible for a malicious or compromised 1Password server to provide dishonest public keys to the user and run a successful attack.”

“We are committed to continually strengthening our security architecture and evaluating it against advanced threat models, including malicious-server scenarios like those described in the research, and evolving it over time to maintain the protections our users rely on,” said Jacob DePriest, 1Password CISO and CIO.

The researchers said their findings can likely be used to target other cloud-based password managers, and well-resourced nation-states may already be doing so. They also said that an informal group of password management vendors has been aware of their research and findings since January 2025, meaning vendors of tools they didn’t review will also have had the chance to find and address any such flaws in their own offerings.

“We cannot exclude the possibility that our attacks were already known to advanced threat actors – after all, we have learned from the Snowden revelations that national security agencies are routinely tasked with penetrating systems like the ones we analyze and are willing to conduct active attacks on targets,” the paper says.

The researchers said they hope to raise the security baseline being offered by cloud-based password managers, by “pushing their vendors to either improve security or make clearer statements about what security their systems actually provide, so that customers can judge (perhaps with the help of expert guides) whether the products meet their requirements or not.”

“My recommendation is still to use a password manager, and I don’t think users should be afraid of cloud-based ones,” Backendal said.