See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/324096476

Analysis on the Security and Use of Password Managers

Conference Paper · December 2017

DOI: 10.1109/PDCAT.2017.00013

CITATIONS READS
22 9,314

4 authors, including:

John Elizarraras Jyh-haw Yeh

Rice University Boise State University
1 PUBLICATION 22 CITATIONS 65 PUBLICATIONS 424 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Jyh-haw Yeh on 27 October 2021.

The user has requested enhancement of the downloaded file.

 Analysis on the Security and Use of Password Managers*
Carlos Luevanos1 , John Elizarraras2 , Khai Hirschi3 , and Jyh-haw Yeh4
1 Dept.of Computer Science, Willamette University, CA
2 North
Star Charter High School, Eagle, ID;
3 Capital High School, Boise, ID
4 Dept. of Computer Science, Boise State University, ID

Abstract— Cybersecurity has become one of the largest The organization of this paper starts with the related works
growing fields in computer science and the technology industry. in Section II. Following that we give a brief rundown and
Faulty security has cost the global economy immense losses. history of password managers. We then discuss each password
Oftentimes, the pitfall in such financial loss is due to the
security of passwords. Companies and regular people alike do manager in their own sections; we will go over Passbolt,
not do enough to enforce strict password guidelines like the Encryptr, and Padlock in that order. Each section contains
NIST (National Institute of Standard Technology) recommends. subsections discussing the details of the password managers,
When big security breaches happen, thousands to millions reported security flaws already found for them, our reported
of passwords can be exposed and stored into files, meaning flaws, a review and critique about the password manager,
people are susceptible to dictionary and rainbow table attacks.
Those are only two examples of attacks that are used to and then some potential solutions to the vulnerabilities. The
crack passwords. In this paper, we will be going over three paper concludes with suggestions of what an ideal password
open-source password managers, each chosen for their own manager should have; we review some features of more
uniqueness. Our results will conclude on the overall security popular password managers and look at the pros and cons of
of each password manager using a list of established attacks both open-source and closed-source password managers.
and development of new potential attacks on such software.
Additionally, we will compare our research with the limited I I . R E L AT E D W O R K
research already conducted on password managers. Finally, we
will provide some general guidelines of how to develop a better The niche of research on password managers has direct
and more secure password manager. ties to that of applied cryptography and penetration testing,
Index Terms— Password Managers, Password Authentication so the body of our work will go back to referencing a mix of
academic papers on the security of more popular password
managers and penetration testing write ups performed by
I . I N T RO D U C T I O N
security auditing teams, although we will reference some
In this paper, we will be discussing three open-source papers that display proof of concept or have contributed to
password managers: Passbolt [24], Padlock [8], and Encryptr the field in some other way. Additionally, we have tried
[6]. We have chosen these three due to each unique quality our best to extend any and all audits performed on our
they carry; we will mention them here once and then bring three researched password managers. The earliest work for
them up again in their separate sections. Passbolt was chosen password managers comes from Luo and Henry in 2003 [20],
for its unique property in that its full potential and benefits are who demonstrated a proof of concept and implementation of
reached when utilized by teams, companies, and closed groups a more effective password manager, compared to Microsoft
of people who trust each other with sensitive information; Passport. In 2005, Halderman et al’s work [19] comprised the
additionally, Passbolt runs on OpenPGP [7], a secure email proof of concept and implementation of a password manager
encryption standard founded by Phil Zimmerman. Padlock in the web-browser, where an example implementation in
was chosen for its use of the Electron and Polymer developer Firefox was given in their work.
environment as well as being a minimalist password manager. Moving on to the trend of security analysis, we are given
Lastly, Encryptr was chosen for its brow-raising qualities; all insight on popular password managers such as LastPass and
data is stored in the cloud, and it uses Crypton, a cryptography Roboform by Li, He, Song, and Akhawe [18]. Silver, Jana,
framework that implements a no-knowledge proof system, a Chen, Boneh, and Jackson [3] made outstanding contributions
somewhat fancier way of saying end-to-end encryption. To to the auto-fill feature found in popular password managers
conclude our paper, we will compare and contrast all the such as LastPass, KeePass, and those implemented in web-
features between these password managers and more popular browsers such as Google Chrome and Safari. They found
ones in order to have a standard of security for this type of critical vulnerabilities that abused the auto-fill feature; such
software in general. attacks include iFrame sweep attacks, password sync exploita-
*This work was partially funded by the NSF REU Software Security Site tion, and injections. Their work would help greatly influence
and NASA ISGC High School Summer Research Experience grants. the policies auto-fill executes. To show some of the root
4 The corresponding author, [email protected] problems as to why password managers are seeming to
become more and more necessary, the work of Gaw and other algorithms if you wish to do so, but we shall not go
Felten [16] would contribute statistical analysis of surveys over technical specifications. Following a mixed response
performed at Princeton University. Gaw and Felten found from the launch of KeePass, it would be a few years until
that participants often reused passwords for less important commercial success for password managers took off; LastPass,
websites and predict this trend would grow as more on-line Dashlane, and Roboform serve as prime examples, with the
accounts accumulate. Participants were found to be ignorant LastPass being the most popular, having a reported 7 million
to the security risks that this trend brings. They also found a users as of 2015 [22]. By the 2010’s it seems as if some kind
feeling of indifference towards the use of password managers. of password manager mania has taken off. Many developers
Looking at the rising trend of cloud computing, the work of have begun focusing their time on password managers, and to
Zhao, Yue, and Sun [14] contributed to vulnerability analysis increase popularity they made their work open-source, giving
of LastPass and Roboform; they were able to detect threats the public a chance to use their product and view their code,
such as credentials being stored in plain-text on cloud servers giving the technical community a chance to shape a password
and offered suggestions to both product makers on how to manager to their liking by giving feedback on what can be
better secure their data and product. added onto or improved, or by exposing vulnerabilities [23].
To conclude the related work we mention Gasti and This method seems to be useful in the fact that it is essentially
Rasmussen [13]; their contributions include a forefront on a free security audit. Of course, the time it would take to
the analysis of password manager database formats, as their report to developers would be much longer.
paper’s title suggests: “On The Security of Password Manager
Database Format.” What Gasti and Rasmussen found was that I V. O P E N - S O U R C E PA S S W O R D M A N A G E R S
despite a number of password managers being different from Now focusing our attention to the bulk of our work, let us
each other, each pretty much used the same database format. recall our three password managers: Passbolt, Padlock, and
They also found several vulnerabilities in each password Encryptr. Each of these password managers is open-source.
manager they investigated. We chose to look at open-source password managers for
multiple reasons. First of all, because they are open-source,
I I I . O V E RV I E W O F PA S S W O R D M A N A G E R S
we can look at the source code ourselves. Additionally, this
A. Quick Rundown also allows us to set up our own servers to test on. Each
First and foremost, we must define what a password man- password manager was chosen for their own unique properties.
ager is. Password managers are programs used to generate, We shall go into technical detail about each one, along with
encrypt, and store passwords for a client-side user. All that pointing out vulnerabilities we have found in each one and
is required of a user is to remember one master password referencing previous vulnerabilities exposed by others.
and user name. It is believed that using such software will
increase security. Typically passwords will be stored on the A. PASSBOLT
local machine itself or on some hosted server. In some cases, Passbolt is an open-source password manager initially de-
they may be hosted on cloud servers to ensure more security veloped by Kevin Muller, Diego Lendoiro, Remy Bertot, and
for the parties involved; i.e., the user and the host of the Cedric Alfonsi, with later work of Passbolt being supported
servers/proprietors of a password manager. There is some by the GitHub community. Passbolt’s core user-base includes
variety in the types of password managers available to the development teams and companies, adopting the philosophy
public; some are built into web browsers such as Google that company password policies can be shoddy or annoying,
Chrome, Mozilla Firefox, Safari, and Microsoft Edge, while which in turn creates a less efficient workspace and perhaps
others serve as standalone programs with the capability of some security vulnerabilities. The Passbolt development team
web-browser integration. Some strictly enforce strong master believes their product can be adopted to ease the process of
passwords while others do not. More notably there are a few sharing passwords among peers and coworkers in an easier
that have integrated multi-factor authentication, which is very and much more secure way [24].
beneficial to security. Given a brief rundown of such software,
let us now delve into some history.
B. History of Password Managers
Not much is known about password managers; the first
successful implementation of an effective password manager
open to the public (to our knowledge), after the work of 1) Overview of Passbolt: Currently, Passbolt only runs in
Luo and Henry [20], was KeePass, developed by Dominik a browser. More specifically, it only runs on Firefox and
Reichl [21], with an initial launch back in 2003. KeePass is Google Chrome. This is reportedly due to them still being in
a minimalist password manager that runs on Windows, Mac the alpha development of Passbolt [25]. Passbolt was written
OS, and Linux, along with unofficial imports to Android, in JavaScript, PHP, and Shell, and it currently uses OpenPGP
IoS, and Blackberry. The latest versions of KeePass use for their encryption standard. Passwords stored in Passbolt
AES-256 bit encryption [40] along with ChaCha20-256 bit are encrypted, and the database used by the client can also be
encryption [41]; however, KeePass does allow you to use encrypted to improve security. However, user names are not
encrypted and are stored in plain text. One existing problem they can steal a user’s master password. See Figure 1 and
for Passbolt includes the use of a bad pseudo-random number Figure 2 for more detail.
generator [25]. There is also no current way to change your
master password or use multi-factor authentication. There
is the option of emailing a copied list of your encrypted
passwords to yourself should the option for email notification
be enabled. Passbolt also boasts the use of a color security
token which should prevent phishing, however, we believe
we have found a way to bypass this feature which we will go
into detail in our discovered flaws section. Another current
flaw reported by the Passbolt team is the predicament of the
client and server trusting all keys; although they admit to this
being a flaw, they wish to fix this error in the future [25].
2) Reported Security Flaws: Found early in the surveil-
Fig. 1: Here we can see the hex value of the color a user’s
lance of Passbolt, it was discovered by Wigginton et al that
security token
the use of the PHPseclib has the potential to default to the
use of ECB encryption [4]. While this isn’t a direct flaw
in Passbolt itself, it is still a flaw to consider. Reported
problems by Passbolt include server integrity problems, DDoS
attacks, server information leaks, key revocation, the potential
of authentication cookies being stolen if SSL is broken, and
the potential to mimic server keys [26]. It should be noted
that Passbolt currently only uses MySQL servers, which have
had reported problems that were recently fixed by Golunski
[2]. Given that Passbolt is only in alpha and developed by
a single team rather than a company, they do not have the
resources to perform a full-stretched security audit; most
vulnerabilities have been found by the developers themselves
or by the GitHub community. The team’s use of cryptographic Fig. 2: By editing this hex value, we can replace the color
functions (by use of OpenPGP) has been reviewed by security with any color of our choice
audit team Cure53 [27]. The team was able to find several
vulnerabilities in the OpenPGP library but we shall omit the
details. One other attack we created was a custom user-script. A
3) Our Discovered Flaws: To begin testing on Passbolt user-script is a custom script that users can install to gain
we opted to not perform any attacks already conducted, so extra functions to websites. Users will typically not look over
we tried some of our own attacks. Before testing, we set up a all of the code in the user-script, so it would be easy to hide
private server using the Hamachi Virtual Private Network two lines of malicious code in a script that looks innocent. We
so that we could as closely as possible simulate a work created a script (see Appendix) that would secretly replace
environment typical to what Passbolt should be used for. The all links that download the Firefox Passbolt extension with
tools involved were all on Kali Linux, with the exception of another random extension. All the links appear to go to the
some attacks written in C# and tested on Windows machines. original website, as seen in Figure 3. Because of the way
It was very easy to see that a key-logger on an unsuspecting Firefox installs extensions, the user will get a pop-up that
user would give us the master password. Passbolt was also says the website itself wants the user to download the fake
found to be susceptible to a clipboard attack; given the extension, making the extension seem trusted (See Figures
generated passwords can be somewhat hard to read, a user 4 and 5). The Passbolt extension itself can’t be modified,
will opt to just copy and paste rather than manually type their however, as Passbolt detects any changes and disables itself
passwords, making strong individual passwords to websites if it finds any. Downloading a fake extension would bypass
and services seem useless if an attacker is successful. this, and since the extension is open-source, mimicking the
Passbolt has a special feature to prevent phishing attacks; extension would be very simple. A proposed method to get
a user will remember a color assigned to them and it will be the user to download this user-script is to hide the code
present when they attempt to log-in. We conducted an attack inside a good user-script. As mentioned previously, the code
on their use of this color key token; by grabbing a live copy would most likely go unchecked. Another method would be
of a user’s session we were able to find the lines of code to hide it in an extension. The extension would then inject
that gave us the color. Simply editing these lines we now the JavaScript into the website.
had an exact copy of what a user would think is the login 4) Review and Critique: Overall, Passbolt was not a very
page, all working as if it were the real thing. With this, an user-friendly password manager and we question the integrity
attacker could potentially perform a phishing attack in which of the product. They have no reported security audits on their
 which is in itself a big security risk.
5) Solutions to Some Vulnerabilities: We suggest that the
Passbolt team develop some features such as typing obfusca-
tion and auto-fill to protect against key-logger and clipboard
attacks. We also encourage the use of other types of servers
besides MySQL. While not all servers are perfect, giving
user options can potentially increase the overall security of
the company using Passbolt. The user-script vulnerability is
something that would be hard for Passbolt to stop, but they
should at least use an HTML content security policy that
would block scripts loaded from an off-line source, most of
the time [12]. We would also recommend some way to either
Fig. 3: Notice how the link to download the extension appears obfuscate the security token or develop some other method
to go to passbolt.com in order to prevent our proposed phishing attack.

B. ENCRYPTR
Encryptr is an open-source password manager initially
developed by Tommy Williams and then bought out by
SpiderOak [28], a company focused on building services that
feature no-knowledge frameworks.

1) Overview of Encryptr: Encryptr is a cross-platform pass-

Fig. 4: Firefox shows the server itself asking you to download word manager, e-wallet, and note-holder written in JavaScript,
the extension HTML, CSS, JSON, and XML. Its encryption standard was
built using the Crypton framework, created by SpiderOak.
Crypton is an open-source framework developed in JavaScript
with a primary goal to store information on a server without
the server ever knowing what is stored [1]. Crypton’s back-
end uses PostgreSQL [36], Redis [37], Node.js [38], and
Docker [39]. Encryption and decryption are assumed in
AES-256 using Galois/Counter Mode. For more specifics,
ElGamal encryption [35] and ECDSA (Elliptic Curve Digital
Signature Algorithm) [34] are used for signature verification,
elliptic curve cryptography is used for key generation, and
such ciphers can be switched for others if a user decides
to do so [1]. The strength of Crypton is the protection of
user data and data sharing, it is the direct belief of end-
to-end encryption that users may feel more secure from
Fig. 5: The link actually downloads an arbitrary extension attackers and the company hosting such a service. Crypton
called NoScript also uses SRP (Secure Remote Password) authentication,
which reportedly limits data compromise, with the only
supposed attack that being brute forcing AES keys [1]. Some
product except a reference to the audit of OpenPGP, and they weaknesses of Crypton include the ability of peer graph
lack many key features other password managers have, which analysis and container access frequency analysis. What we
brings into question why they would even release an alpha mean by the former statement is that user-names are stored in
version of their product. Additionally, their demo page is plain text, so it is possible for database records to be analyzed
quite shoddy and the intended use of the password manager in order to find connections between users and perform some
seems to often lead attackers to use cunning phishing attacks intelligence gathering, leading to more potential attack vectors
and DDoS attacks. The design also increases the risk of an [1]. As for the latter, containers in Crypton can sometimes be
attack on the main administrator, since they are at the core of created, updated, and accessed deterministically, which can
how Passbolt should be used by a company. It should also be lead to a potential brute force attack. However, this can be
noted there is no enforcement of a strong master password, remedied using a strong password [1]. One last note about
Crypton and other SpiderOak applications is the use of the obfuscation should user’s create their own passwords instead
clipboard, which is reported by them to be safer than just of generating one.
typing all your information.
2) Reported Security Flaws: It should be noted in this C. PADLOCK
section that no official security audit of Encryptr has been Padlock is a minimalist, open-source password manager
performed. Furthermore, to our knowledge no security audit developed by Martin Kleinschrodt using the Electron and
has been performed on other SpiderOak services. Only two Polymer frameworks, so like Passbolt and Encryptr, it was
official security audits were performed and published on all written in JavaScript, HTML, and CSS, and for the most
Crypton. The main issues reported by Leviathan Security part, all code is available to the public. Padlock is multi-
Group include [10]: platform and can be used on Windows, MacOS, Android,
1) An account’s public key is not verified against the iOS, and some time in the future, Linux.
decrypted private key. This could result in a user
encrypting something that cannot be decrypted.
2) The public signing key is not verified against the private
signing key.
3) A container by the name of containerNameHMacKey
is not verified before decryption, so the server could
replace it with a different known container and encrypt 1) Overview of Padlock: Similar to Passbolt and Encryptr,
a new symmetric key to the user’s public key. Padlock uses a copy/paste function to quicken the process of
A look into the report by Least Authority gave us some using one’s passwords. As noted earlier, this leads to very big
more insight into the security of Crypton which includes [11]: security problems. One of the notable features of Padlock
is that the application automatically logs you out of your
1) Server information forgery: attackers with access to the
vault in one minute if there is no user activity detected. This
server can overwrite and forge data on a user’s account.
feature can be changed to a maximum of ten minutes or can
2) Guessable private keys: an attacker with server access
even be disabled if the user wishes to do so. Padlock does
can grab copies of cipher-text and read the plain-text.
include its own password generator, however, its weakness
3) An attacker with server access can disclose the en-
is that 7 character passwords with at least one uppercase
cryption key, essentially making all container contents
letter, lowercase letter, and special character are considered
available to them.
very strong by Padlock. By the 2017 NIST standards, these
Having read both reports we noticed plenty of the attacks generated passwords are not considered secure [29].
included DDoSing. While not the largest security threat to 2) Reported Security Flaws: Surprisingly, the creator of
worry about, it can still cost companies quite a bit of money Padlock has a repository for penetration testing of his own
and reputation. The reports failed to include auditing web- application. Furthermore, the penetration team Cure53 was
based attacks such as XSS attacks (Cross-Site Scripting), hired to do even more extensive testing. The reports detail
CSRF attacks (Cross-Site Request Forgery), Man-In-the- some of the following vulnerabilities:
Browser attacks, and SQL injections.
• Tap-jacking [9].
3) Our Discovered Flaws: Our reported findings of vul-
• Exposed authentication tokens during API requests,
nerabilities for Encryptr include:
leading to Man in the Middle attacks [9].
1) High security threat with clipboard attacks (copying is • Permanent DoS attack on mobile devices: an attacker
the easiest way to transfer passwords from Encryptr to with server access can increase the number of iterations,
a form). essentially making the CPU do a job it cannot do, and
2) Susceptible to key-loggers when typing your master ultimately having to make the user reset Padlock if
password or when transferring passwords from Encryptr they want to use the application on their phone again.
to a form without copy-paste. However, resetting Padlock will delete all information
4) Review and Critique: Encryptr was the most minimal stored [9].
of the three open-source password managers reviewed by • DoS email attacks [9].
far; it was incredibly simple, it could be used on almost all 3) Our Discovered Flaws: Since Padlock uses the clip-
platforms and did not require the use of an email, yet you board like Passbolt and Encryptr, along with no auto-fill
could still retrieve the same data from other devices. One feature, it was easy to discover that Padlock is susceptible
critique of Encryptr is that after some further investigation it to clipboard attacks and key-loggers. Additionally, we wrote
was discovered some code is still obfuscated. It is also noted a script (see Appendix) that would be able to reset a user’s
that there is no strict enforcement of strong passwords and account, all that is required is the user click a button, thus
generated passwords have a default length of 12 characters. deleting all their saved passwords and information. The attack
5) Solutions to Some Vulnerabilities: Like for Passbolt, involves a user installing a script or extension which would
we recommend SpiderOak implement an auto-fill feature for then inject JavaScript into the Padlock web page. For testing
passwords and credit-card information, along with typing purposes, we did this using our own personal server, but we
believe it can be easily adapted to any other server using servers is also a nice feature that can potentially increase
Padlock. We also used a user-script (defined in subsection security for users who know what they are doing, but this
IV-A.3) to manipulate a vulnerability in Padlock. Padlock has can be detrimental when attackers gather knowledge of targets
an on-line dashboard where users can change what devices using their own servers. It should be noted that the use of
have access to their account, as well as reset their data–with end-to-end encryption by Encryptr, i.e. Crypton, is a very
no password. To abuse this fault, we made a script that would desirable feature, especially for those who want the utmost
reset the user’s data the instant they logged into the dashboard. privacy they can get. Such no-knowledge features were even
It should be noted that Padlock does try to prevent this. It uses commented on by Edward Snowden, who is pushing for end-
content security policies that block off-line scripts [12]. This to-end encryption to become a more standard feature in cloud
method will break certain user-script managers (extensions storage [30].
that install and inject the user-scripts) but some managers
are able to bypass the security policy. This is something B. Weaknesses of Open-Source Password Managers
Padlock can’t fix but they should ask for more verification, Being open-source allows people to find vulnerabilities,
such as a password, to reset all of the user’s data. It is just but not everyone will report the security problems they
two clicks to clear all of the passwords and devices on an find. An attacker can keep a vulnerability they found secret
account. It should be noted that if you don’t reset the client, and use it in a future attack. Additionally, quite a few
your passwords are safe, but if the cloud was the only place open-source password managers do not have many features
where the passwords were stored, this method will make the that strengthen security like more well-known, closed-source
passwords unrecoverable. Another script (see Appendix) we password managers do. In order for an open-source product to
made would revoke all of the devices, meaning that the user be successful, it must have a strong group of supporters that
would lose access to the cloud on their devices. This isn’t a review the code and make suggestions, and a development
huge issue, as the user can just reconnect, but it demonstrates team that listens to feedback and works quickly and diligently.
Padlock’s vulnerability to script attacks.
4) Review and Critique: Padlock lived up to its name of C. Strengths of Closed Source Password Managers
being a minimalist password manager that got the job done Closed source password managers have the benefit of
and we were quite pleased with the initial security audits that keeping their code hidden from potential attackers. This
were reported on the application. Overall it was easy to use means that an attacker usually won’t be able to see the
and the ability to use a custom server was a nice addition. code and exploit vulnerabilities found in it. It also means
However, we did not like the minimum security standards of duplicating the password manager is harder, so certain attacks,
generated passwords, nor was there any strict enforcement of like a fake extension, wouldn’t be as effective. We would
strong master passwords. also like to point out some of the desirable features that
5) Solutions to Some Vulnerabilities: Similar to Passbolt closed-source password managers such as LastPass have.
and Encryptr, we suggest the creation of an auto-fill feature as LastPass includes features such as auto-fill, and two-factor
well as typing obfuscation to prevent key-logger and clipboard authentication using your phone, or fingerprint. LastPass
attacks. Additionally, we suggest that the reset feature be also includes the feature of passwords only being local,
removed, as loss of all passwords in one swoop can be easy; meaning they are stored only on the machine [31]. Closed
whether they be by someone with direct access to the local source password managers also include features such as
machine or by some social-engineering attack. secure file and password sharing, tracking history of what
sites were logged-in, and reports of the current strength of
V. C O N C L U S I O N : W H AT M A K E S A your passwords. Such features were included in password
PA S S W O R D M A N A G E R M O R E S E C U R E managers such as LastPass, Dashlane, and Roboform [32].
In this section, we will review the strengths and weaknesses
of all the password managers we looked at in comparison D. Weaknesses of Closed Source Password Managers
with the strengths and weaknesses of more popular password The main weakness of a closed source password manager is
managers in order to envision what a more secure password the proprietor behind it. The user has to trust that the company
manager would look like. is securely storing their passwords, since the user doesn’t
know the details of how their passwords are being stored
A. Strengths of Open-Source Password Managers and processed. The proprietor is also responsible for security
Looking back at Passbolt, Encryptr, and Padlock, one of the updates, and they will usually have fewer people looking over
greatest strengths they commonly share is the fact that they each line of code than an open-source project would. The
are open-source. This allows for consumers to examine the users have to trust that the owner of the password manager
code and report any vulnerabilities and bugs themselves to is responsible and active in providing security updates. Fur-
quicken the refinement process. This method of putting trust thermore, while it is not feasible to claim that closed-source
into the consumer eases some of the burdens for developers password managers are at risk of facing more attacks, the
and, of course, saves money; however, this does come at implications behind a security breach are detrimental; larger
the cost that bugs and vulnerabilities may be found at a closed-source managers risk losing massive amounts of assets
much slower pace. The option of users setting up their own and user data since they are larger targets with more resources
to take from. Such an example can be seen from LastPass [14] Rui Zhao, Chuan Yue and Kun Sun, ”Vulnerability and Risk
in the first quarter of 2017, an exploit found by a Google Analysis of Two Commercial Browser and Cloud Based Password
Managers,” http://inside.mines.edu/ ruizhao/Docs/Papers/bcpmsPAS-
researcher revealed a flaw that could have let attackers exploit SAT2013 Jour.pdf
the LastPass browser extension [33]. [15] Sonia Chiasson, P.C. van Oorschot, and Robert Biddle, ”A Usability
Study and Critique of Two Password Managers,” Proceedings of the
E. Theoretical Design for a Good Password Manager 15th Conference on USENIX Security Symposium, 15(1), 2006.
[16] Shirley Gaw and Edward W. Felten, ”Password Management Strategies
A good password manager would prioritize security over for Online Accounts,” Proceedings of the Second Symposium on Usable
ease of use. Firstly, the password manager would be open- Privacy and Security, pp. 44-55, 2006.
source to ensure that anyone who uses it can know what [17] Scott Standridge, ”Password Management Applications and Practices,”
https://www.sans.org/reading-room/whitepapers/bestprac/password-
it is doing to protect their privacy. While a long password management-applications-practices-36755
may be annoying to the user, the master password must be [18] Zhiwei Li, Warren He, Devdatta Akhawe and Dawn Song, ”The
strong. If the master password is weak to brute force attacks, it Emperor’s New Password Manager: Security Analysis of Web-based
Password Managers,” 23rd USENIX Security Symposium (USENIX
brings down the security of the whole password manager. For Security 14), pp. 465-479, 2014.
this reason, the password manager would require a complex [19] J. Alex Halderman, Brent Waters and Edward W. Felten, ”A convenient
master password that meets the 2017 NIST standards. We method for securely managing passwords,” Proceedings of the 14th
international conference on World Wide Web, pp. 471-479, 2005.
would also suggest to add an auto-fill function to the password [20] H. Luo and P. Henry, ”A common password method for protection
manager. This would prevent clipboard and keylogger attacks of multiple accounts,” 14th IEEE Proceedings on Personal, Indoor and
if implemented correctly. Our approach to auto-fill would be Mobile Radio Communications, Vol. 3, pp. 2749-2754, 2003.
[21] Dominik Reichl, ”KeePass - The Open Source Password Manager,”
to press a browser extension button to activate the auto-fill http://keepass.info/
function on the current site, much like certain other password [22] Sarah Perez, ”LogMeIn Acquires Password Man-
managers already do. We would also lock the user’s vault agement Software LastPass For $110 Million,”
http://social.techcrunch.com/2015/10/09/logmein-acquires-password-
after a specified period of inactivity set by the user (no longer management-software-Lastpass-for-110-million/
than two hours). The user would then need to type in their [23] Katherine Noyes, ”10 Reasons Open Source Is Good for Business,”
master password again before using the password manager. http://www.pcworld.com/article/209891/10 reasons open source is good
for business.html
For cloud storage, we would suggest using a no-knowledge [24] ”Passbolt | Credits,” https://www.passbolt.com/credits
approach, similar to Encryptr, to ensure that the server and [25] ”Passbolt | FAQ,” https://www.passbolt.com/faq
any server-side attackers don’t have access to any password. [26] ”Passbolt | Authentication,” https://www.passbolt.com/help/tech/auth
[27] Mario Heidrich, ”pentest-report openpgpjs.pdf,”
We would allow the use of a custom server, but force the use https://cure53.de/pentest-report openpgpjs.pdf
of HTTPS. We would try to make every step as automated [28] Tommy Williams, ”Encryptr Now an Official SpiderOak Prod-
as possible to increase usability. This would include setting uct,” https://devgeeks.tumblr.com/post/132849662534/encryptr-now-an-
official-spideroak-product
up a custom server, since all of the password managers we [29] Paul Grassi, James Fenton, Elaine Newton and William Burr, ”NIST
tested had a very difficult setup process. Special Publication 800-63B.”
[30] Anthony Ha, ”Edward Snowden’s Privacy Tips: Get Rid
REFERENCES Of Dropbox, Avoid Facebook And Google,” http://social.
techcrunch.com/2014/10/11/edward-snowden-new-yorker-festival/
[1] Cam Pedersen and David Dahl, ”Crypton: Zero-Knowledge Application [31] ”Features | LastPass,” https://lastpass.com/features/
Framework,” 2014. [32] Neil J. Rubenking, ”The Best Password Managers of 2017,” PC
[2] David Golunski, ”MySQL-Maria-Percona-PrivEscRace-CVE-2016- Magazine, https://www.pcmag.com/article2/0,2817,2407168,00.asp
6663-5616-Exploit,” https://legalhackers.com/advisories/MySQL-Maria- [33] Colin Lecher, ”LastPass security flaw could have let hack-
Percona-PrivEscRace-CVE-2016-6663-5616-Exploit.html ers steal passwords through browser extensions,” The Verge,
[3] David Silver, Suman Jana, Dan Boneh, Eric Chen and Collin Jackson, 2017, https://www.theverge.com/2017/3/22/15023062/lastpass-security-
”Password Managers: Attacks and Defenses,” 23rd USENIX Security flaw-passwords
Symposium (USENIX Security 14), pp. 449-464, 2014. [34] Don Johnson, Alfred Menezes and Scott Vanstone, ”The
[4] P.I.E. Staff, ”Choosing the Right Cryptography Library for your Elliptic Curve Digital Signature Algorithm (ECDSA),”
PHP Project: A Guide - Paragon Initiative Enterprises Blog,” http://cs.ucsb.edu/ koc/ccs130h/notes/ecdsa-cert.pdf
https://paragonie.com/blog/2015/11/choosing-right-cryptography- [35] Jaspreet K. Grewal, ”ElGamal: Public Key Cryptosystem,”
library-for-your-php-project-guide http://cs.indstate.edu/ jgrewal/steps.pdf
[5] Luke Graham, ”Cybercrime costs the global economy $450 billion: [36] ”PostgreSQL: About,” https://www.postgresql.org/about/
CEO”, http://www.cnbc.com/2017/02/07/cybercrime-costs-the-global- [37] ”Introduction to Redis,” https://redis.io/
economy-450-billion-ceo.html [38] Node js Foundation, https://nodejs.org/en/about/
[6] Spideroak, ”Free, Secure Password Manager - No Knowledge End-to- [39] ”What is Docker,” 2015, https://www.docker.com/what-docker
End Encryption,” https://spideroak.com/personal/encryptr [40] Joan Daemen and Vincent Rijmen, ”AES proposal: Rijndael,” 1999,
[7] ”OpenPGP,” http://openpgp.org/ http://www.cs.miami.edu/home/burt/learning/Csc688.012/rijndael/ rijn-
[8] Maklesoft, ”Padlock - A Minimalist Password Manager,” dael doc V2.pdf
https://padlock.io [41] Daniel J. Bernstein, ”ChaCha, a variant of Salsa20,” Workshop
[9] M. Heiderch, ”Pentest-Report Padlock.io,” Record of SASC, Vol. 8, pp. 3-5, 2008, http://ai2-s2-pdfs.
https://padlock.io/docs/padlock-pentest-1604.pdf s3.amazonaws.com/2ea9/7a1597dfa8d74c6e544fb4709532ef587c69.pdf
[10] Leviathan Security Group, ”Crypton Security Audit,”
http://roselabs.nl/files/audit reports/Leviathan SpiderOak Crypton.pdf
[11] Enterprises, Least Authority, ”Least Authority Performs Security Audit
For SpiderOak | Least Authority,” https://leastauthority.com
[12] Fondeo Inc.,”Content Security Policy CSP Reference & Examples,”
https://content-security-policy.com/
[13] Paolo Gasti and Kasper B. Rasmussen, ”On the Security of Password
Manager Database Formats,” Lecture Notes in Computer Science,
Computer Security – ESORICS 2012, pp. 770-787, 2012.
 APPENDIX

1 // Changes all of the links that download the Firefox extension

2 var links =
document.querySelectorAll("a[href='https://www.passbolt.com/download/firefox']"); //
gets all links meant to download the Passbolt firefox extension
3 for (var i = 0, len = links.length; i < len; i++) { // go through all the links
4 var link = links[i];
5 link.setAttribute('onclick', "location.href='https://goo.gl/2WjK9u';return false;");
//makes the href change to our page when the link is clicked. This allows us to hide
the real location of the link since it only changes where it goes after you click the
link.
6 }
Listing 1: Script that will make all links that download the Firefox Passbolt Extension change to download NoScript (An
arbitrarily chosen extension)

1 // Clicking reset data link

2 var links = document.querySelectorAll("a[href='.?action=resetdata']"); //gets the link to
reset data
3 for (var i = 0, _len = links.length; i < _len; i++) {
4 var link = links[i];
5 link.click(); // clicks the link to reset
6 }
7
8 // Clicking buttons
9 setTimeout(function() {
10 var buttons = document.getElementsByTagName('button'); //gets all buttons on page
11 for (var j = 0; j < buttons.length; j++) { // Loop through all buttons
12 if (typeof buttons[j].click === "function") { //checks to see if button.click is a
function
13 buttons[j].click(); // clicks the button
14 }
15 }
16 }, 1); // Times out to allow webpage to show buttons after clicking link

Listing 2: Script for Padlock that resets all of the data once the user logs into their dashboard

1 // Submits the form that revokes all devices on Padlock Cloud

2 var forms = document.forms; // gets all forms on page
3 for (var i = 0, len = forms.length; i < len; i++){ // Loop through all forms
4 forms[i].submit();// submit all of the forms found
5 }
Listing 3: Script that revokes all devices connected to a Padlock Cloud once the user logs into the dashboard

For documentation on these scripts, go to https://github.com/iblacksand/vulnerabilitydocumentation. It features the full scripts,

a keylogger, and a clipboard reader. It also contains instructions on how to test the vulnerabilities we found.

View publication stats