Search Mailing List Archives

Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[liberationtech] Draft checklist for choosing tools

Karl Fogel kfogel at
Sun Dec 30 10:40:19 PST 2012

bobalice at writes:
>Getting someone who gets their security advice from Forbes to evaluate and
>select tools for secure private communications is probably non-trivial,
>but worthwhile.
>This is, hopefully, the start of a practical general checklist.
>Comments and suggestions would be appreciated. Happy holidays!

That's a useful checklist, thanks.  Are you posting it anywhere (I mean,
on a wiki or web page, besides this mailing list)?

My only comments are below, in quoted context.

>It may be daunting to evaluate and choose a tool for your needs. As
>individual circumstances differ, you should choose a tool that best fits
>your situation, taking into account responsibility and risk.
>For a suitable tool, answers to the checklist questions should all be
>`yes' or not applicable.
>Do your own research, but do reach out to the community (for example,
>liberationtech) for assistance and to confirm your conclusions.
>Is the installer or program (binary) digitally signed?
>Why: A binary should carry a valid and trusted digital signature to ensure
>that the correct file is received, whether downloaded from the Internet or
>obtained from a friend in the field. The digital signature provides a link
>between the signer, ideally, the creator of the file and its checksum. If
>the file with a valid signature turns out to be compromised or contains a
>backdoor, the signer can be asked to account for the discrepancy, even if
>it turns out their computer was compromised (more likely) or an adversary
>is capable of forging the digital signature (very notable, but less
>likely). A good digital signature should use a current checksum/digest
>(SHA1/SHA256/SHA512 and not MD5/CRC32, which are very weak) and a
>sufficiently long public key (2048 bits or higher). In the case of a
>certificate authority-based digital signature, the chain of certificates
>should be plausible. For example, some obscure entity located in a distant
>land should not be asserting that the signer is Mozilla.
>*Open source*
>Is the source code available?
>Why: Having source code available for anyone to examine offers the
>security community and other interested observers the ability to inspect
>the implementation for bugs, flaws, and vulnerabilities. While adversaries
>may use the opportunity to find exploitable vulnerabilities, the security
>community has the same chance to do so. For closed source tools, it is
>possible that only the authors and adversaries have access to the source.

Strictly speaking, the code just has to be "visible source" in order for
others to examine it.  

Visible source is a subset of open source.  In both cases, anyone can
see the code.  In the open source case, they are also free to share it,
modify it, and redistribute modified versions.  Obviously, that helps
with other requirements on your checklist, such as the "Sustainability"
requirement given later (where I've commented), and in general I think
open source is a very good thing, of course.  But it's not actually
necessary for examinable code.

>Is the source code signed?
>Why: Code that does not carry valid digital signatures can be altered in
>transit. The version of code that you receive may not be the version that
>is used to build the program. (Also, see next item.) Digital signatures
>associated with commits helps with accountability and integrity.
>Can you use the source code to produce a functionally identical version of
>the tool?
>Why: Although source code might be available, that code may not actually
>be used to produce the tool that you obtained. For example, a compromised
>binary might be built from a combination of the clean public source code
>and a backdoor. If a functionally identical version of the tool cannot be
>produced from the provided source code, this situation, among others, may
>be more difficult to detect and prevent as you cannot simply build your
>own version to use. NB: The standard is currently a functionally identical
>version of the program. The act of making a complex program bit-for-bit
>identical is an active area of work due to differences in, for example,
>compile times and dates.

The above two requirements also do not require open source.

>Does the project welcome code contributions?
>Why: If bug fixes cannot be contributed, the incentive for individuals in
>the security community to examine the source can be reduced. In addition,
>this type of open source project may rely solely on the developer to fix
>bugs and be more dependent on the actual good behavior of the developer.

Even this doesn't really require open source.  (However, people are
probably more likely to contribute bugfixes to a project that allows
them to redistribute the code under an open source license.)

Users are almost never in a position to evaluate this property, and much
as I hate to admit it, there are open source projects of high code
quality that are not very good about accepting code contributions.
Anyway, one can be good at accepting bug reports but bad at accepting
code contributions; those are semi-independent qualities.

I guess I'm not sure this one belongs on the list.  It's a useful
property when evaluating a project's open source health, but it might
only have second-order relevance for security and code quality.

>*Threat model*
>Does the creator define the assumptions that underlie the tool's operation?
>Why: It is necessary to evaluate if the tool is right for your use case
>and to make an informed decision with relation to its shortfalls and
>benefits. The threat model, risks, benefits, warnings, and implementation
>details should be available and complete in both plain and technical
>language. In particular, what does the tool not protect against (for
>example, a smartphone that already has malicious apps)?
>Are you advised (realistically) of the risks and benefits?
>Why: It is necessary to know how the tool may fail. Does the tool creator
>take into account likely, but not yet documented, capabilities of
>adversaries (for example, partial network view in case of Tor)? Although
>some risks and failure cases may not be immediately apparent (even to the
>tool creators), known issues (or issues apparent by inspection) should be
>made obvious in plain language and presented alongside expected benefits.
>Is the tool secure by design?
>Why: The tool creator may be compelled, by legal means or otherwise, to
>provide user data or other assistance to an adversary. Is the tool
>designed in such a way to minimize the threat arising from this situation
>(for example, through end-to-end encryption or by having any sensitive
>data revealed to the tool creator or their infrastructure)? If the tool
>does not work for some reason, does it fail in a safe manner (for example,
>no network access for a VPN as opposed to allowing connections without
>Are you sufficiently protected from the future?
>Why: Suppose your counterparty's computer is compromised or is seized
>sometime in the future by an adversary who has also captured an encrypted
>version of your communications. If your tool does not have perfect forward
>security, your past communications might be decrypted using information
>from your counterparty's system. Likewise, if the cryptographic components
>used in the chosen tool are not strong enough to resist expected advances
>in computational power, the communications might be decrypted at some
>point in the future as an expected development.
>Are the algorithms and components used plausible?
>Why: While it may be necessary to consult the security community (for
>example, liberationtech) about this point, improper choices in parameters,
>cryptographic components, and system design can result in an unsafe tool.
>Is this best tool for your use case, despite any drawbacks?
>Why: It is essential to examine the tool landscape and make an informed
>decision for your use case. In the end, the decision might be to modify
>the use case. For example, one may decide to use a known insecure tool
>(for example, telephone) in conjunction with an one-time pad to avoid a
>false sense of security.
>Is there an active developer base to address bugs and changes in the
>security landscape?
>Why: Critical bugs may be present despite code review and good design
>decisions. Bugs may also be introduced by time, for example, through an
>increase in computational power. These bugs need to be addressed in a
>timely fashion. In addition, changes to an adversary's capabilities may
>require cryptographic upgrades (longer key lengths, obfuscation). Without
>active maintenance, tools that were once safe may become unsafe over time.
>Are the project's finances sustainable?
>Why: If the tool relies on infrastructure (for example, RTP proxies for
>VoIP), the cost needs to be covered somehow (donations, user payments,
>which may not be anonymous, by selling user data, etc.). In addition,
>finding and switching to a different tool in the field can be difficult.

Sustainability is a stronger argument for open source than the
transparency arguments were, I think.  

Open source projects are more sustainable because the code can live
independently of the bad decisions of any particular maintainer.
Someone might be good at coding but bad at business or organizational
decisions -- that's okay, someone else can eventually fork the code and
maintain it in a more sustainable way.  This has happened before
(remember that Mozilla is a fork of the old Netscape Navigator source
code, which was open sourced toward the end of its business life).

Similarly with the "active developer base": open source projects can be
picked up by anyone -- in other words, if a tool is widely used and open
source, then it will not die, because *someone* will pick up
maintenance.  This affects the switching cost argument: you're less
likely to have to switch away from an open source tool, because it's
less likely to become unmaintained (I'm writing these words in the same
open source mailreading editor I started using 20 years ago).

So I think there are good arguments for preferring open source tools,
but they're not quite the arguments you gave.

These are minor nits, of course.  I think it's a great checklist, and
really hits all the important points.


More information about the liberationtech mailing list