Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 10 Oct 2014 00:12:28 +0400
From: Solar Designer <>
Subject: liability (was: Re: Thoughts on Shellshock and beyond)

I ended up writing a lengthy message (this one), but I am unsure if it's
a good idea to have this topic discussed once again (such discussions
had already occurred on other mailing lists years ago).  In fact, that's
the main point I am making - while I've just spent/wasted some time on
writing the below, maybe we should stop right here?  So if anyone has
something new or some important historical references to add, please feel
free to post, but I'd rather not see us digress into (I think) mostly
irrelevant analogies in financial markets, with even more irrelevant
detail on the French trader (referring to Sven's other posting here).

I mention some paywalled articles below.  If anyone has URLs to free
copies of those, please post.

On Thu, Oct 09, 2014 at 11:11:34AM +0200, Sven Kieske wrote:
> so at least when you're making money of software you should
> be responsible for this software.

That's tricky.  Is an Open Source project that accepts donations, sells
CDs/DVDs, or/and runs ads on the website "making money"?  What if they
also offer related paid services or even occasionally sell commercial
licenses to the same software?  Would they be liable e.g. for up to all
payments they ever received (or more?), even if 99.9% of the users never
paid anything?  That may easily put them "out of business", or
discourage them from starting the project in the first place.

Of course, you can hope to reduce undesired effects of a new law by
careful wording, listing categories of software it does or/and does not
apply to, etc.  However, getting the legal system involved at all is a
huge risk... yet you'd like to use it to reduce risk elsewhere?  The
legal system is already akin to an over-engineered software program, and
you're proposing to make it even more complex (more buggy, and requiring
more resources to run).  What's worse, you don't get to write that
"program", and you can't replace it on your "computer" with some
alternative (short of moving to another jurisdiction, and even that
option might disappear if the law becomes universally accepted).  You
can request a "feature", and if the powers that be listen, they'll
implement that "feature" in some arbitrary way that you might not like,
yet all of us would be stuck with it.  In my opinion, this is extreme
danger, possibly way beyond the risk from software vulnerabilities (to
the extent that risk could be reduced by such measures).  Indeed, these
are different types of risks, so a direct comparison of this sort may
only make sense in specific contexts (e.g., effect on a country's
economy or on people's quality of life analyzed in some specific way).

I am not saying I am strictly against this approach, although that is my
current stance given the (frankly) rather limited impact that software
vulnerabilities actually have on us so far despite of being widespread.
(I think the negative impact of introducing liability for software
vulnerabilities might well be broader.)  What I am saying is that it's a
really tough tradeoff, and that in my opinion anyone who feels confident
about it is either wrong in being so confident or has values different
from mine.

> that's also not just my opinion (and I didn't invent these
> thoughts), some credit has to go out to mr Schneier who
> you might happen to know ;)
> see:

Oh, this is definitely an old topic, much older than 2003, although that
was a year when it was discussed more.

There were two related articles published in IEEE Security & Privacy,
2003 vol.1, Issue 1 - January-February:

Two Views on Security Software Liability: Let the Legal System Decide
Daniel J. Ryan , Law Offices of Daniel J. Ryan
pp. 70-72

Two Views on Security Software Liability: Using the Right Legal Tools
Carey Heckman , Darthmouth College
pp. 73-75

Here's the abstract for the first one of these:

"Rather than use the product liability screwdriver as a chisel, why not
consider a package of more effective tools. Corporations and individuals
that market software despite knowledge of software security flaws should
face criminal prosecution as well as civil lawsuits with punitive
damages. Perhaps bounties should be available for the first to discover
and establish the existence of a security flaw. Publishers should be
required to post to the Web and otherwise publicize promptly patch
availability. The software equivalent of an Underwriters Laboratories
should establish and constantly improve security-related standards and
testing protocols. It should be made readily apparent whether a program
has passed and at what level. Prospective customers should be educated
and encouraged to insist on software that has passed. Stronger software
security is important. Software developers and publishers must do
better. But product liability is not the right legal tool for the job."

To me, this suggests a much more limited use of the legal system: with
liability only for continued marketing of software with security flaws
already known to the vendors.  Even that, though, is incompatible even
with current best practices, where vendors of any large piece of
software almost constantly have some vulnerabilities (and fixes for
them) in the pipeline, yet they don't stop marketing the software during
those periods (which would be almost all the time, and would ruin their

I haven't read the actual articles.  Anyone found them non-paywalled?

This presentation appears to argue in favor of strict liability and
backs this up with some theory:

Open Source is briefly mentioned on slide 10.  I think it fails to
consider (at least) the negative impact extra regulation will have on
innovation.  To limit liability risks, a software product would need to
be compliant with current best practices, not with potentially better
practices that are not yet widely accepted and would necessarily be
considered at least good by courts.

[ While we're at it, I think the same problem - negative impact on
innovation - applies to another old idea, insurance for IT-related risks
including security compromises:

Worse, besides innovation this would also discourage diverse setups -
e.g., use of less popular operating systems, which insurers have no
statistics on and don't want to bother with - even though these setups
might actually be safer from actually occurring attacks (at least the
non-targeted ones) precisely due to the diversity. ]

Some maybe-relevant historical publications:

Liability and computer security: Nine principles
Ross J. Anderson
Proceedings of Third European Symposium on Research in Computer Security
Brighton, United Kingdom
November 7-9, 1994
pp. 231-245

Here's the abstract:

"The conventional wisdom is that security priorities should be set by
risk analysis. However, reality is subtly different: many computer
security systems are at least as much about shedding liability as about
minimising risk. Banks use computer security mechanisms to transfer
liability to their customers; companies use them to transfer liability
to their insurers, or (via the public prosecutor) to the taxpayer; and
they are also used to shift the blame to other departments (we did
everything that GCHQ/the internal auditors told us to). We derive nine
principles which might help designers avoid the most common pitfalls."

A related thought is that if vendors would be liable for software
vulnerabilities, they'd want to proactively turn those into
non-vulnerabilities as far as the law is concerned, by proactively
transferring liability onto other parties (their customers, etc.)
Right now, they can do it in EULA.  If they would be legally barred from
that, they might find other ways.  For example, rather than say "We
disclaim all liability" in the EULA, the program could have e.g.
execution of arbitrary code from e-mail messages described as intended
behavior that the user authorizes.  OK, perhaps that would impact the
vendor's sales, if their competitors don't do the same.  But possibly
they'll find smarter ways, including those that would use "computer
security mechanisms" (as the above abstract says) - just not those
protecting the users, but rather those helping transfer liability away
from the vendor.

Some older publications I happened to find, apparently by lawyers:

Product Liability and Software
Michael C. Gemignani
Rutgers Computer & Tech. L.J.
p. 173

The Business Lawyer
Vol. 26, No. 4 (April 1971)
pp. 1081-1094

Unfortunately, these are also paywalled.

Most discussion on infosec mailing lists that I can find now is from the
period 2000 to 2005, although I vaguely recall seeing this topic brought
up in 1990s as well.


Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.