Device-Bound vs. Synced Credentials:
A Comparative Evaluation of Passkey Authentication
Andre B
¨
uttner
a
and Nils Gruschka
b
University of Oslo, Gaustadall
´
een 23B, 0373 Oslo, Norway
Keywords:
Passkeys, FIDO2, Passwords, Authentication, Security.
Abstract:
With passkeys, the FIDO Alliance introduces the ability to sync FIDO2 credentials across a user’s devices
through passkey providers. This aims to mitigate user concerns about losing their devices and promotes
the shift toward password-less authentication. As a consequence, many major online services have adopted
passkeys. However, credential syncing has also created a debate among experts about their security guarantees.
In this paper, we categorize the different access levels of passkeys to show how syncing credentials impacts
their security and availability. Moreover, we use the established framework from Bonneau et al.’s Quest to
Replace Passwords and apply it to different types of device-bound and synced passkeys. By this, we reveal
relevant differences, particularly in their usability and security, and show that the security of synced passkeys
is mainly concentrated in the passkey provider. We further provide practical recommendations for end users,
passkey providers, and relying parties.
1 INTRODUCTION
The primary authentication method for online ser-
vices is still a password, the same as it was at the be-
ginning of the Web. Nowadays, further authentication
factors like OTP codes often accompany it, but the un-
derlying security and usability issues, like phishing or
forgotten passwords, remain unchanged. However, a
recent trend towards password-less authentication can
be observed. This development is driven by big-tech
companies such as Apple, Google, and Microsoft. To-
gether with many other organizations, they formed the
Fast Identity Online (FIDO) Alliance which envisions
a future without passwords
1
. The FIDO2 standard en-
ables the user to register a cryptographic public key
with a service and authenticate by proving the pos-
session of the secret credential—now widely referred
to as passkey.
Initially, the standard only considered credentials
stored on one device, such as a security key or a sin-
gle smartphone or computer. However, due to the
concern about losing the authenticator device, FIDO2
experienced a relatively low adoption by both ser-
a
https://orcid.org/0000-0002-0138-366X
b
https://orcid.org/0000-0001-7360-8314
1
https://fidoalliance.org/overview/ (Last accessed:
2024-12-16)
vice providers and end users. This has led to the
paradigm shift towards synced credentials that can be
exchanged between devices. Consequently, an on-
line account can still be accessed from another de-
vice when one device gets broken or lost. Synced
passkeys require a passkey provider, an additional en-
tity in the FIDO2 ecosystem that provides access to
credentials across different devices, just like password
managers. Most major password managers have al-
ready implemented passkey management and, there-
fore, also function as passkey providers.
The decision to pivot to synced passkeys has
caused quite some confusion and criticism. Since
passkeys can be transmitted between devices, there is
a higher risk of remote attacks. Furthermore, passkey
providers may become a single point of failure in case
they get compromised. It has been shown in previ-
ous instances where password managers have been
hacked, such as the severe breach of LastPass in 2022
(Toubba, 2022), that this is an actual threat. More-
over, another question is whether passkeys mitigate
the risk of account lockout if an authenticator is lost.
So far, little research has been done on the se-
curity and usability of passkeys, and even less has
considered the differences between the different types
of passkeys. In this paper, we aim to systematize
the knowledge and concerns regarding passkeys and
present the observations from our exploratory re-
Büttner, A. and Gruschka, N.
Device-Bound vs. Synced Credentials: A Comparative Evaluation of Passkey Authentication.
DOI: 10.5220/0013380600003899
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 11th International Conference on Information Systems Security and Privacy (ICISSP 2025) - Volume 2, pages 651-659
ISBN: 978-989-758-735-1; ISSN: 2184-4356
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
651

search. Our contribution can be summarized as fol-
lows:
• We categorize the different access levels of
passkey credentials introduced by passkey
providers and highlight how these affect security
and availability.
• We systematically compare passwords, device-
bound passkeys, and synced passkeys using the
framework presented in (Bonneau et al., 2012)
and show that the security of synced passkeys is
mainly concentrated in the passkey providers.
The remaining sections of this paper are organized as
follows. In Section 2, related work is described. Sec-
tion 3 addresses different passkey access levels and
Section 4 entails our comparative evaluation of differ-
ent passkey types. Finally, Section 5 includes the dis-
cussion, and Section 6 summarizes our findings and
gives an outlook for future work.
2 RELATED WORK
Regarding security, the FIDO2 standard has been
analyzed using formal methods, uncovering poten-
tial weaknesses in the underlying protocols (Barbosa
et al., 2021; Guan et al., 2022; Bindel et al., 2023).
Some articles have examined specific vulnerabilities,
such as the possibility of intermediary entities that can
exploit FIDO2 extensions (Xu et al., 2021; B
¨
uttner
and Gruschka, 2022) and attacks against QR-code-
based registration of a passkey (Kim et al., 2024).
Others have discovered local attacks against FIDO2,
bypassing browser or OS protection measures (Yadav
and Seamons, 2023; Mahdad et al., 2024). We extend
the work above by highlighting the role of passkey
providers as a crucial part of the FIDO2 ecosystem.
To the best of our knowledge, academic re-
search has not yet investigated the security of synced
passkeys. However, there are contributions from the
industry with suggestions on how to evaluate risks
regarding passkeys. For example, there is a blog
post (Iozzo and Mroz, 2023) where a threat model for
passkeys is proposed. While it refers to relevant as-
pects such as server- and client-side credential steal-
ing, it is limited and only makes suggestions for rely-
ing party implementations. Furthermore, in a talk at
the Authenticate 2023 conference hosted by the FIDO
Alliance, another threat model was described that
considers several risks of passkeys related to creden-
tial compromise, recovery, and sharing (Saxe, 2023).
The usability of FIDO2 authentication is another
important topic that has been addressed by several
researchers. Some studies have analyzed the usabil-
ity of smartphones as FIDO2 authenticators (Owens
et al., 2021; W
¨
ursching et al., 2023); however, at a
time when they were only stored on a single device.
Furthermore, user perceptions of security keys have
been studied (Lyastani et al., 2020). Moreover, us-
ability aspects have been analyzed with a particular
focus on using FIDO2 in enterprise contexts (Kep-
kowski et al., 2023). Only one study we found also
considered the usability and deployability aspects of
synced passkeys. Here, CISOs and FIDO2 experts
were interviewed regarding their thoughts on using
passkeys in companies, showing remaining concerns,
such as account recovery and technical issues (Lassak
et al., 2024).
In summary, our literature review has identified
a lack of research on passkeys and the influence of
passkey providers and their features. Since passkey
providers substantially change the way FIDO2 cre-
dentials are accessed, our paper highlights important
factors for the usability, deployability, and security
risks of passkeys.
3 ACCESS LEVELS OF
PASSKEYS
With a passkey provider, users can distribute their cre-
dentials to different devices, share them with other
users, and manually back them up. However, vari-
ous risks can occur depending on how restricted the
access to a user’s passkeys is. Based on our ob-
servations, we define four different access levels of
passkeys that are (1) bound to one device, (2) synced
between the devices of a user, (3) shared with other
users, and (4) exported in plaintext and thus accessi-
ble outside the scope of a passkey provider (see Fig-
ure 1). One can assume that a lower access restriction
for a passkey leads to a larger attack surface, thus in-
creasing the risk level. In the following, we describe
each of the different access levels and their impact on
passkey credentials.
3.1 Device-Bound Passkey
A device-bound passkey corresponds to the initial
concept behind FIDO2 authentication. It is a cre-
dential that is only stored on the device where it was
generated. Therefore, it does not require a passkey
management tool. Since it can be protected by secure
hardware, it provides the highest security guarantees,
making it nearly impossible for anyone to obtain a
private key. Even for an attacker with physical ac-
cess to the device, it will be difficult to steal a user’s
ICISSP 2025 - 11th International Conference on Information Systems Security and Privacy
652

Exported
Shared
Synced
Device-Bound
Risk
Single-User
Context
Multi-User
Context
External Scope
Passkey Provider Scope
Device Scope
Figure 1: Access levels of passkey credentials. The estimated risk is indicated by the colors green (low-risk), yellow (low- to
medium-risk), orange (medium- to high-risk ), and red (high-risk).
device-bound passkey, provided it is protected by user
verification mechanisms, e.g., a PIN or fingerprint.
3.2 Synced Passkey
Synced passkeys are generated on one device and dis-
tributed to others through a passkey provider. This en-
sures that passkey credentials remain accessible when
one of the devices is damaged or lost. Since passkeys
can be transmitted from one device to another, they
cannot be isolated in secure hardware. This increases
the chances of a remote attacker capable of installing
malware on the victim’s device, stealing the passkey
from the memory.
Further, passkeys must be protected when synced
between the user’s devices using end-to-end encryp-
tion. Another risk is that a user’s passkey provider
account may get compromised. It is, therefore, cru-
cial that a user protects their account with secure au-
thentication methods. Thus, new devices should be
authenticated to ensure an attacker cannot access the
keys. Google, for instance, requires the user to en-
ter the Google Password Manager PIN or the screen
lock from an Android device that already has ac-
cess to the passkeys before syncing credentials (De-
sai, 2024). We noticed, however, that some passkey
providers only require a master password, which pro-
vides access to both the passkey provider account and
the database.
3.3 Shared Passkey
Passkey providers may also allow users to share their
credentials with other users. We observed this on Ap-
ple’s password manager and most third-party passkey
providers. This drastically increases the attack sur-
face of passkeys as the original owner may lose con-
trol over their credentials. As soon as a passkey is
shared, one must assume this operation is irreversible.
Even if a passkey provider lets a user limit or revoke
access to a passkey by other users, it will not stop a
malicious user from extracting the passkey out of the
passkey provider. Therefore, this feature should be
used carefully and only with trusted users. Particu-
larly in enterprises, one must consider internal threats
and ensure that former employees do not keep access
to their passkeys.
When sharing passkeys, users also rely on protect-
ing all user accounts that can access them. An attacker
may be able to determine whether one of those ac-
counts is only weakly protected and exploit it accord-
ingly. Another concern could be social engineering,
where a user is tricked into sharing a passkey with an
account controlled by an attacker.
3.4 Exported Passkey
Many passkey providers allow users to export their
passkeys to back them up or migrate them to an-
other provider. However, as soon as passkeys leave
the scope of the passkey provider, protection mea-
sures such as access control and database encryp-
tion become ineffective. Thus, the backup file should
be password-encrypted, which is offered by many
passkey providers, to protect its content from an at-
tacker who gains access. This is especially relevant
when storing the backup in the cloud. In any case, the
export functionality should be used carefully.
The FIDO Alliance has further published a work-
ing draft for a protocol for secure credential exchange
between different services. This would enable pass-
word and passkey management tools to migrate cre-
dentials with lower effort by the user and with higher
protection (FIDO Alliance, 2024).
Device-Bound vs. Synced Credentials: A Comparative Evaluation of Passkey Authentication
653

4 EVALUATION OF PASSKEYS
We apply the framework presented in (Bonneau et al.,
2012) to evaluate passkeys and compare different us-
ability, deployability, and security aspects.
4.1 Passkey Types
Authenticators for device-bound credentials, which
correspond with the original FIDO2 concept, can be
divided into platform and roaming authenticators. A
platform authenticator is embedded into the client
device from which a user authenticates to a relying
party. Such a device can be a computer, a tablet, or
a smartphone. The operating system on the client
device provides an API that enables the creation of
cryptographic keys, which are generated and stored
in a secure hardware module. In contrast, roaming
authenticators are separate devices and physically iso-
lated from the client device. Common examples are
security keys that can be plugged into the client via
USB or connected via NFC. Mobile devices can also
be used as roaming authenticators in cross-device au-
thentication contexts where the mobile device con-
nects to the client via Bluetooth.
Since the shift to passkeys, credentials can also
be synced between different devices using a passkey
provider. In this case, the FIDO Alliance dis-
tinguishes between first- and third-party passkey
providers
2
. First-party providers are essentially the
same as built-in password managers, i.e., already pro-
vided through the operating system or browser. At
the time of writing, this mainly includes Apple’s
and Google’s respective password managers. Note
that passkeys stored in Windows Hello cannot yet
be synced to date. External applications supporting
passkey management, such as Bitwarden, ProtonPass,
or KeePassXC, are considered third-party providers.
4.2 Methodology
We compare the four types of passkeys described
above with each other and in addition to legacy pass-
words, i.e., traditional passwords memorized by the
user. Since the intention behind passkeys is to replace
passwords, we do not consider multi-factor authenti-
cation. Moreover, we only address the concepts and
intended features of passkeys. Implementation vul-
nerabilities on the relying party, client, or authentica-
tor are not considered since they can eventually affect
any authentication scheme. We further assume using
2
https://passkeys.dev/docs/reference/terms/ (Last ac-
cessed: 2024-12-16)
TLS-encrypted traffic between the client and the rely-
ing party.
The framework in (Bonneau et al., 2012) has been
applied in various studies to compare different au-
thentication methods (Dressel et al., 2019; Lyastani
et al., 2020; Arias-Cabarcos et al., 2021; Kunke et al.,
2021). It proposes a list of 25 benefits that may (or
may not) apply to particular authentication schemes,
including eight benefits for usability, six for deploya-
bility, and eleven for security. We adopt their evalua-
tion of legacy passwords as a baseline. For evaluating
the passkeys, we discuss each of the benefits carefully
to decide whether the respective passkey type offers,
partially offers, or does not offer it.
4.3 Results
Table 1 summarizes our assessment of the different
benefits of passwords and each passkey type. In the
following, we describe our reasoning for the evalua-
tion in detail.
4.3.1 Usability
Regarding usability, all types of passkeys offer more
benefits than passwords. The characteristics of be-
ing Memorywise-Effortless and Scalable-for-Users
are inherently given because passkeys are created and
stored by the authenticator, and the user does not need
to remember a secret for each of the services that are
accessed. This applies to all four types of passkeys.
One could argue that a user must remember a PIN or
master password for the authenticator devices. How-
ever, this has been replaced by biometric authentica-
tion for modern computers, smartphones, or security
keys. Therefore, memory effort is negligible com-
pared to remembering a password for each website.
Passkeys offer scalability since registering for mul-
tiple services does not require the user to remember
multiple secrets, unlike passwords.
While passwords are Nothing-to-Carry, they do
not offer the benefit of being Physically-Effortless
since the user has to type them for every new ses-
sion. A roaming authenticator is, by definition, a sep-
arate device from the client and does not offer the
former benefit. Regarding the physical effort, roam-
ing authenticators must be connected via USB, NFC,
or Bluetooth. Hence, they do not provide the lat-
ter benefit either. In contrast, platform authenticators
and synced passkeys offer both benefits because they
are accessed on the client device from which the user
wants to sign in. Moreover, they do not require physi-
cal effort because the device unlocking mechanism is
used, usually a fingerprint scan or face recognition, to
which the user is accustomed.
ICISSP 2025 - 11th International Conference on Information Systems Security and Privacy
654

Table 1: Comparison of passwords and different passkey types based on the framework in (Bonneau et al., 2012).
Usability Deployability Security
Memorywise-Effortless Scalable-for-Users Nothing-to-Carry Physically-Effortless
Easy-to-Learn
Efficient-to-Use Infrequent-Errors Easy-Recovery-from-Loss
Accessible
Negligible-Cost-per-User Server-Compatible Browser-Compatible
Mature
Non-Proprietary
Resil.-to-Physical-Observation Resil.-to-Targeted-Impersonation Resil.-to-Throttled-Guessing Resil.-to-Unthrottled-Guessing
Resil.-to-Internal-Observation
Resil.-to-Leaks-from-Other-Verifiers Resil.-to-Phishing Resil.-to-Theft No-Trusted-Third-Party Requiring-Explicit-Consent
Unlinkable
Legacy Password
*
# # # G# # G# # # # # #
Device-Bound
Platform Authenticator # G#
Roaming Authenticator # # G# G# # G# G#
Synced
First-Party Provider G# G# G# G# #
Third-Party Provider # G# G# G# #
= offers benefit; G# = partially offers benefit; # = does not offer benefit
* from (Bonneau et al., 2012)
We consider both types of device-bound passkeys
to be Easy-to-Learn since they do not require any spe-
cific knowledge by a user other than a simple inter-
action. We assume synced passkeys with first-party
providers to be less easy to learn. In some cases,
the users must configure passkey syncing explicitly,
for instance, when using the Google Password Man-
ager. This is not obvious for average users, and it is
also not trivial from which devices the passkeys can
be accessed. Third-party providers can be even less
understandable because, in addition to the challenges
mentioned above that arise with first-party providers,
they require an additional application or browser ex-
tension to be installed. Also, third-party providers
have to be enabled explicitly on mobile devices in the
device settings. When testing different providers, we
observed that they provide different user interfaces,
which can confuse the user, particularly when using
different passkey providers simultaneously.
Once the user understands the concept, we assume
passkeys to be Efficient-to-Use. For security keys,
it could be shown, e.g., in two-factor authentication
contexts, that they perform significantly better than
other methods (Reese et al., 2019; Reynolds et al.,
2020). The only exception is roaming authenticators,
such as phones, that must be connected via Bluetooth.
In this case, an unavoidable delay occurs because a
user first needs to scan a QR code displayed on the
client device, and the roaming authenticator subse-
quently needs to connect to it. Therefore, we rate
device-bound roaming authenticators only as partially
offering this benefit.
The benefit Infrequent-Errors is provided when
using platform authenticators or synced passkeys
since the user only needs to confirm the authentication
procedure. There may be minor issues during bio-
metric verification or other types of device unlocking.
However, we consider this to be insignificant com-
pared to the problems with typing passwords. Similar
to the previous benefit, we argue that roaming authen-
ticators may be error-prone when using Bluetooth de-
vices, particularly when scanning QR codes.
Account recovery is one of the most controversial
aspects of passkeys (Lyastani et al., 2020; Kuchhal
et al., 2023). Although different proposals for re-
covery strategies for device-bound FIDO2 credentials
have been evaluated (Frymann et al., 2020; Kunke
et al., 2021) they have not been widely adopted.
Thus, device-bound passkeys do not provide Easy-
Recovery-from-Loss. The FIDO Alliance has ad-
dressed this issue with synced passkeys, which are
supposed to minimize the risk of account lockout.
However, its effectiveness ultimately depends on how
many of the user’s devices store the key. Also, there is
a chance that a user loses or forgets their master pass-
word. Since the database of a credential manager is
usually encrypted with the master password, it might
not be possible for the user to recover the data. Still,
we conclude that the risk of losing access is reduced,
and therefore, the benefit is partially offered for both
Device-Bound vs. Synced Credentials: A Comparative Evaluation of Passkey Authentication
655

types of synced passkeys.
4.3.2 Deployability
Regarding the deployability benefits, we consider
passkeys nearly as good as passwords. The first two
benefits in this category Accessible and Negligible-
Cost-per-User depend only on the client and the
passkey device or provider.
We assume all types of passkeys to be accessible
to anybody who is otherwise capable of using pass-
words. Regarding costs, roaming authenticators such
as security keys can be expensive, ranging in price
from 30$ to over 100$. On the other hand, smart-
phones can be used as roaming authenticators and do
not involve any additional costs. Thus, depending on
the device type, we conclude this benefit to be par-
tially offered. In contrast, platform authenticators and
synced passkeys can be used without any cost. Some
third-party providers may require passkey features to
be purchased, but we found that most commercial
providers offer passkey support for free.
The remaining deployability benefits depend es-
sentially on the FIDO2 standard and are equal for
all types of passkeys. According to (Bonneau et al.,
2012), the benefit Server-Compatible requires com-
patibility with text-based passwords on the server
side. While passkeys do not fulfill this, we interpret
this benefit rather as how expensive it is for a server to
provide this authentication method. We rate this ben-
efit partially offered due to its wide support by many
services
3
and the availability of various WebAuthn li-
braries
4
. The benefit Browser-Compatible is offered
by passkeys since the WebAuthn standard (Kumar
et al., 2021) is implemented by all major browsers.
Because of the aforementioned broad support both on
the server and client, we also consider passkeys to
offer the benefit Mature. Furthermore, passkeys are
Non-Proprietary since FIDO2 is an open standard and
can be implemented by anyone.
4.3.3 Security
Concerning security, all passkey types offer the
first four benefits. They are Resilient-to-Physical-
Observation because observing a user authenticate
with a passkey will not disclose the underlying private
key used to sign the challenge. Similarly, passkeys
are Resilient-to-Targeted-Impersonation since back-
ground knowledge will not bring an attacker any
3
According to 1Password, more than 200 services sup-
ported passkeys in December 2024. See https://passkeys.
directory/ (Last accessed: 2024-12-01)
4
https://passkeys.dev/docs/tools-libraries/libraries/
(Last accessed: 2024-12-13)
closer to obtaining the passkey since it is a
possession-based type of authentication. Fur-
thermore, they are resilient to both Throttled-
and Unthrottled-Guessing due to the underlying
challenge-response protocol using modern digital sig-
nature algorithms.
For the benefit Resilient-to-Internal-Observation,
we regard the protection of the private key as highly
relevant. Therefore, it is only offered entirely by
device-bound authenticators. Roaming authenticators
and platform authenticators isolate the private key in
secure hardware and prevent potential malware on
the client from stealing a key from memory. Synced
passkeys cannot provide this protection entirely since
the key must be accessible to passkey provider soft-
ware to be shared between devices. However, since
passkeys are end-to-end encrypted during transmis-
sion, they are still much more difficult to steal than
intercepting passwords.
Passkeys are Resilient-to-Leaks-from-Other-
Verifiers because relying parties only store a public
key and identifiers that do not correlate across
services. If this information is leaked, it does not help
an attacker in any way to break into a user’s account
on another service.
Phishing resistance is one of the main advantages
of passkeys over passwords. This is achieved through
origin verification by the client and by an authen-
ticator signing challenges only for a specific rely-
ing party. All types of passkeys implement this and
consequently offer the benefit Resilient-to-Phishing.
Passkeys are also Resilient-to-Theft as the device they
are stored in is usually locked by a PIN or biometrics
with a limited number of attempts.
Device-bound passkeys are inaccessible to third-
party services since they are stored only on one
device. Therefore, they offer the benefit No-
Trusted-Third-Party. This is different for synced
passkeys. Although the FIDO Alliance denotes cer-
tain providers as first-party, in terms of security, we
consider all passkey providers as third-party services
in relation to the user. The providers must be trusted
to operate as intended and protect the user’s creden-
tials. Many aspects must be considered here, such as
account protection on the passkey provider service,
storage and encryption of the database, and who else
is eligible to access the data. These factors can vary
for each passkey provider and each user.
Passkey authentication, as with passwords, re-
quires some user interaction and fulfills the benefit
Requiring-Explicit-Consent. Finally, passkeys also
offer the last benefit Unlinkable. In contrast to pass-
word authentication, where users tend to reuse a
password on different services, the information for
ICISSP 2025 - 11th International Conference on Information Systems Security and Privacy
656

passkey authentication stored by different relying par-
ties cannot be linked to each other. The WebAuthn
standard (Kumar et al., 2021) even enables username-
less authentication through discoverable credentials,
thus further improving the user’s privacy.
5 DISCUSSION
In this section, we reflect on our passkey evaluation,
discuss limitations of our analysis, refer to the re-
lationship between passkey providers and password
managers, and offer some practical recommendations.
5.1 Comparison of Passkey Types
In terms of usability, roaming authenticators have sev-
eral drawbacks. Platform authenticators, in contrast,
offer the most usability benefits because they are inte-
grated into the users’ devices. However, both types
have in common that they are not easy to recover.
Since credentials are device-bound, a user cannot re-
cover them if the authenticator device gets lost. In-
stead, users must configure recovery methods in their
accounts, which requires much effort. As this caveat
has been highlighted several times as a major obstacle
to adopting FIDO2, it is highly relevant to end users.
Synced passkeys are less likely to get lost, provided
they are stored on several devices or backed up by the
passkey provider or the user. Their main downside is
that users might not be accustomed to them, and their
functionality is less transparent than passwords.
All types of passkeys are similar regarding their
deployability since most benefits are related to the
FIDO2 protocol specification. The cost of security
keys is the only factor that makes roaming authenti-
cators less favorable in comparison. As mentioned
earlier, the benefit of being compatible with servers
is nearly given since several libraries are available.
Therefore, one can say that passkeys perform almost
as well as passwords in terms of deployability.
Passwords lack many security benefits, while
passkeys perform significantly better. Device-bound
passkeys fulfill all of the benefits, according to our as-
sessment. Synced passkeys provide many more ben-
efits than legacy passwords because of asymmetric
cryptography, eliminating the risk of guessing creden-
tials and phishing. The security of synced passkeys
is mainly concentrated in the passkey providers as
a trusted third-party service. If they are vulnerable
or act maliciously, a user’s credentials may be com-
pletely unprotected, depending on how passkey cre-
dentials are accessed (as discussed in Section 3).
5.2 Limitations
In this paper, we present a comparison of differ-
ent passkey types. We emphasize that this research
is a relatively high-level qualitative analysis. The
main goal is to provide an overview of the differ-
ent passkey types and highlight their differences. In
particular, the benefits from Section 4 that we eval-
uate as partially offered require a more fine-grained
distinction, e.g., of the respective authenticator de-
vice or a specific passkey provider. The usability
aspects should be reviewed in actual usability stud-
ies to empirically assess the users’ perception of the
different passkey types. Furthermore, the security
analysis does not cover the implementation details of
passkey providers, like their authentication methods
or database storage. However, our evaluation shows
that many security problems with passwords are irrel-
evant when using passkeys and that we need to focus
on specific security aspects of passkey providers.
5.3 Passkey Provider vs. Password
Manager
Most password managers now also support passkeys
and treat both passkeys and passwords the same. Con-
sequently, many aspects of passkeys mentioned in this
paper apply analogously to passwords stored in pass-
word managers. Both types of credentials can be
synced across devices, shared with others, and ex-
ported. In the study from (Lassak et al., 2024), some
participants even mentioned that passkey providers
are essentially better password managers. In both
cases, they interact with a credential management tool
instead of entering their credentials. However, a com-
mon risk is that password and passkey providers can
become a single point of failure, potentially compro-
mising several user accounts through a single breach.
Nonetheless, we have focused mainly on the dif-
ference between device-bound and synced passkeys.
We argue that it is essential to distinguish between
passwords and passkeys because their core concepts
have different security properties. Passwords can not
be protected through secure hardware like device-
bound passkeys. Moreover, passwords stored in a
password manager are often manually created by the
user and can thus be guessed by an attacker (Pear-
man et al., 2019). However, the most important dif-
ference is that passwords will always be vulnerable to
phishing, even when generated randomly. Users must
always be cautious when entering their passwords or
copy-pasting them from their password manager into
a website.
Device-Bound vs. Synced Credentials: A Comparative Evaluation of Passkey Authentication
657

5.4 Recommendations
For end users, it is important that passkey credentials
are protected and made available. Given the secu-
rity benefits, we recommend using passkeys on on-
line services instead of passwords whenever possible.
However, passkey providers should be chosen mind-
fully based on device compatibility to ensure creden-
tials are synced on multiple devices. Furthermore,
users should consider a provider offering strong au-
thentication methods to ensure attackers cannot ac-
cess their database. They should choose a strong mas-
ter password to prevent any credential leaks even in
case of an account compromise or a breach into the
passkey provider. Credential sharing should be used
only when necessary. Ideally, as soon as the other
users do not need to access the respective account
anymore, the passkey should be replaced by a new
one to prevent misuse. If users manually back up their
credentials by exporting them, they should store them
only offline or encrypt them.
Passkey providers should offer robust access con-
trol measures to prevent unintended database access.
We observed that some credential managers use the
same password both for authenticating to the creden-
tial manager account and for encrypting the database.
If this password is leaked, e.g., by a phishing attack,
the adversary gains full access to the user’s creden-
tials. Consequently, service providers should, at best,
make the users create a different password for the
credential manager account or encourage multi-factor
authentication. In our evaluation, we also noticed the
diverse user interfaces of the passkey providers when
authenticating to a relying party with a passkey. We
presume that this may confuse users and that there is
a need for improvements on this end.
Relying parties do not have much influence on
the users’ interaction with passkey providers. How-
ever, they can decide whether they allow their users
to authenticate with device-bound or synced passkeys
through device attestation. If device-bound creden-
tials are used, they should support the user in set-
ting up proper recovery methods. We also find that
passkeys are often offered as an alternative authenti-
cation method to the password rather than a replace-
ment. Relying parties should not wrongly convey a
sense of increased security. Instead, they should still
recommend that their users set up a secondary authen-
tication factor or apply risk-based authentication so
the password does not undermine account security de-
spite setting up a passkey.
6 CONCLUSION
The FIDO Alliance’s proposal to sync passkeys
across different devices has increased the adoption of
passkeys by online services but also received criticism
from security experts for its potential risks.
To address this, we derived different access levels
and their impact on the attack surface of passkeys and
conducted a comparative evaluation of different us-
ability, deployability, and security benefits for device-
bound and synced passkeys. We argue that synced
passkeys mitigate credential loss; however, only with
limited guarantees as it depends on whether passkey
syncing is configured on several devices or whether
the passkey provider backs them up. Moreover, we
show that all passkey types provide more security
benefits than passwords.
Our findings support claims that synced passkeys
are less secure than device-bound ones. However, the
range between secure and insecure passkeys varies
widely depending on their implementation and us-
age. Since many users are unwilling to adopt device-
bound passkeys, more effort should be made to help
end users use synced passkeys most advantageously.
Thus, we emphasize the need for strong authentica-
tion for passkey provider accounts, cautious use of
credential-sharing, and secure storage of backups.
For future work, we propose analyzing the user’s
perception of synced passkeys and whether this ac-
tually increases their motivation to adopt them. Be-
yond this, implementation vulnerabilities of passkey
provider applications should be analyzed to evaluate
the protection of synced credentials in more detail.
REFERENCES
Arias-Cabarcos, P., Habrich, T., Becker, K., Becker, C., and
Strufe, T. (2021). Inexpensive brainwave authentica-
tion: new techniques and insights on user acceptance.
In 30th USENIX Security Symposium (USENIX Secu-
rity 21), pages 55–72.
Barbosa, M., Boldyreva, A., Chen, S., and Warinschi, B.
(2021). Provable security analysis of FIDO2. In Ad-
vances in Cryptology–CRYPTO 2021: 41st Annual
International Cryptology Conference, CRYPTO 2021,
Virtual Event, August 16–20, 2021, Proceedings, Part
III 41, pages 125–156. Springer.
Bindel, N., Cremers, C., and Zhao, M. (2023). FIDO2,
CTAP 2.1, and WebAuthn 2: Provable security and
post-quantum instantiation. In 2023 IEEE Sympo-
sium on Security and Privacy (SP), pages 1471–1490.
IEEE.
Bonneau, J., Herley, C., Oorschot, P. C. v., and Stajano, F.
(2012). The Quest to Replace Passwords: A Frame-
work for Comparative Evaluation of Web Authentica-
ICISSP 2025 - 11th International Conference on Information Systems Security and Privacy
658

tion Schemes. In 2012 IEEE Symposium on Security
and Privacy, pages 553–567.
B
¨
uttner, A. and Gruschka, N. (2022). Protecting FIDO
Extensions Against Man-in-the-Middle Attacks. In
International Workshop on Emerging Technologies
for Authorization and Authentication, pages 70–87.
Springer.
Desai, C. (2024). More users can now save passkeys in
Google Password Manager. https://blog.google/
technology/safety-security/google-password-
manager-passkeys-update-september-2024/ (Last
accessed: 2024/12/13).
Dressel, T., List, E., and Echtler, F. (2019). SecuriCast:
zero-touch two-factor authentication using WebBlue-
tooth. In Proceedings of the ACM SIGCHI Symposium
on Engineering Interactive Computing Systems, pages
1–6.
FIDO Alliance (2024). FIDO Alliance Publishes
New Specifications to Promote User Choice and
Enhanced UX for Passkeys - FIDO Alliance.
https://fidoalliance.org/fido-alliance-publishes-
new-specifications-to-promote-user-choice-
and-enhanced-ux-for-passkeys/ (Last accessed:
2024/12/13).
Frymann, N., Gardham, D., Kiefer, F., Lundberg, E., Man-
ulis, M., and Nilsson, D. (2020). Asynchronous Re-
mote Key Generation: An Analysis of Yubico’s Pro-
posal for W3C WebAuthn. In Proceedings of the 2020
ACM SIGSAC Conference on Computer and Commu-
nications Security, pages 939–954.
Guan, J., Li, H., Ye, H., and Zhao, Z. (2022). A formal
analysis of the FIDO2 protocols. In European Sympo-
sium on Research in Computer Security, pages 3–21.
Springer.
Iozzo, V. and Mroz, K. (2023). Passkeys - Threat
modeling and implementation considerations.
https://www.slashid.dev/blog/passkeys-security-
implementation/ (Last accessed: 2024/07/18).
Kepkowski, M., Machulak, M., Wood, I., and Kaafar, D.
(2023). Challenges with passwordless FIDO2 in an
enterprise setting: A usability study. In 2023 IEEE
Secure Development Conference (SecDev), pages 37–
48. IEEE.
Kim, D., Kim, S., Ryu, G., and Choi, D. (2024). Ses-
sion Replication Attack Through QR Code Sniffing
in Passkey CTAP Registration. In IFIP International
Conference on ICT Systems Security and Privacy Pro-
tection, pages 294–307. Springer.
Kuchhal, D., Saad, M., Oest, A., and Li, F. (2023). Evalu-
ating the Security Posture of Real-World FIDO2 De-
ployments. In Proceedings of the 2023 ACM SIGSAC
Conference on Computer and Communications Secu-
rity, pages 2381–2395.
Kumar, A., Jones, J., Hodges, J., Jones, M., and Lundberg,
E. (2021). Web Authentication: An API for accessing
Public Key Credentials - Level 2.
Kunke, J., Wiefling, S., Ullmann, M., and Lo Iacono,
L. (2021). Evaluation of Account Recovery Strate-
gies with FIDO2-based Passwordless Authentica-
tion. In Open Identity Summit 2021, pages 59–70.
Gesellschaft f
¨
ur Informatik e.V., Bonn.
Lassak, L., Pan, E., Ur, B., and Golla, M. (2024). Why
Aren’t We Using Passkeys? Obstacles Companies
Face Deploying FIDO2 Passwordless Authentica-
tion. In 33rd USENIX Security Symposium (USENIX
Security 24), pages 7231–7248, Philadelphia, PA.
USENIX Association.
Lyastani, S. G., Schilling, M., Neumayr, M., Backes, M.,
and Bugiel, S. (2020). Is FIDO2 the Kingslayer of
User Authentication? A Comparative Usability Study
of FIDO2 Passwordless Authentication. In 2020 IEEE
Symposium on Security and Privacy (SP), pages 268–
285.
Mahdad, A. T., Jubur, M., and Saxena, N. (2024). Breach-
ing security keys without root: Fido2 deception at-
tacks via overlays exploiting limited display authen-
ticators. In Proceedings of the 2024 on ACM SIGSAC
Conference on Computer and Communications Secu-
rity, CCS ’24, page 1686–1700, New York, NY, USA.
Association for Computing Machinery.
Owens, K., Anise, O., Krauss, A., and Ur, B. (2021). User
Perceptions of the Usability and Security of Smart-
phones as FIDO2 Roaming Authenticators. In Sev-
enteenth Symposium on Usable Privacy and Security
(SOUPS 2021), pages 57–76. USENIX Association.
Pearman, S., Zhang, S. A., Bauer, L., Christin, N., and Cra-
nor, L. F. (2019). Why people (don’t) use password
managers effectively. In Fifteenth Symposium on Us-
able Privacy and Security (SOUPS 2019), pages 319–
338, Santa Clara, CA. USENIX Association.
Reese, K., Smith, T., Dutson, J., Armknecht, J., Cameron,
J., and Seamons, K. (2019). A Usability Study of
Five Two-Factor Authentication Methods. In Fifteenth
Symposium on Usable Privacy and Security (SOUPS
2019), pages 357–370.
Reynolds, J., Samarin, N., Barnes, J., Judd, T., Mason,
J., Bailey, M., and Egelman, S. (2020). Empirical
Measurement of Systemic 2FA Usability. In 29th
USENIX Security Symposium (USENIX Security 20),
pages 127–143.
Saxe, D. (2023). Authenticate 2023 The New Passkey
Threat Model. https://www.youtube.com/watch?v=
1TRViLIxAMU (Last accessed: 2024/07/18).
Toubba, K. (2022). 12-22-2022: Notice of Security
Incident. https://blog.lastpass.com/posts/notice-of-
recent-security-incident (Last accessed: 2024/12/16).
W
¨
ursching, L., Putz, F., Haesler, S., and Hollick, M. (2023).
FIDO2 the rescue? Platform vs. roaming authentica-
tion on smartphones. In Proceedings of the 2023 CHI
Conference on Human Factors in Computing Systems,
pages 1–16.
Xu, P., Sun, R., Wang, W., Chen, T., Zheng, Y., and Jin, H.
(2021). SDD: A trusted display of FIDO2 transaction
confirmation without trusted execution environment.
Future Generation Computer Systems, 125:32–40.
Yadav, T. K. and Seamons, K. (2023). A Security and
Usability Analysis of Local Attacks Against FIDO2.
arXiv preprint arXiv:2308.02973.
Device-Bound vs. Synced Credentials: A Comparative Evaluation of Passkey Authentication
659