[Cryptography] I don't get it.

Sandy Harris sandyinchina at gmail.com
Wed Apr 16 09:42:50 EDT 2014


On Tue, Apr 15, 2014 at 6:19 PM, Bowness, Piers <piers.bowness at rsa.com> wrote:

> The ability to detect these types of issues is beyond a typical compiler's job; it cannot
> infer intent from the code being converted from language to machine code.
>
> Static analysis tools, OTOH, ...

There are plenty of such tools available. Modern compilers are often better
at this than older ones, especially if you enable the right options. Compiling
with two or more different compilers is a good check. So is running the code
through one of lint(1)'s descendants.

Some of this could be built into source code management, release
build scripts or make files. Require a clean compile with certain
options and a clean run of lint before a submission is accepted,
for example.

Some concerns can be raised with just a simple grep or a bit of perl
code to do your own checks. My first cut at this, for example, is a
< 10-line program that finds > 3800 files in the Linux kernel where
the word "switch" occurs more often than "default". Nowhere near
all of these will be actually be problematic, but they are worth a look.


More information about the cryptography mailing list