> GCC would be a lot harder and people would be looking for it.
   GCC yes, but what about Visual Studio? The LLVM which ships with XCode?
   Javac?
   Those are much jucier targets than say, Debian's GCC. If I worked at
   the NSA, I'd make backdooring Visual Studio my 20% project.
   If anybody with Snowden cache access who is reading this could do me a
   quick favor and grep for 'compiler,' and then publish the output, I'd
   really appreciate it.
   Thanks,
   R
   On Tue, Feb 11, 2014 at 2:17 PM, Troy Benjegerdes <[1]hozer@hozed.org>
   wrote:

     All the 'NDA'/proprietary/confidential information that goes with
     chip designs
     provide plenty of cover to insert backdoors.
     GCC would be a lot harder and people would be looking for it.
     But your USB chip, graphics card, hard drive, or two factor
     authentication
     token, on the other hand...
     The chinese are probably even copying the subverted chip designs
     without
     even knowing it's there.

   On Tue, Feb 11, 2014 at 02:47:01PM -0700, Kelly John Rose wrote:
   > I could see them more easily subverting chip designs themselves then
   trying
   > to subvert the entire compiler ecosystem.
   >
   >
   > On Tue, Feb 11, 2014 at 2:05 PM, CypherPunk <[2]cypherpunk@cpunk.us>
   wrote:
   >
   > >
   > > On 02/11/2014 01:32 PM, Rich Jones wrote:
   > > > In all of the Snowden docs that have been released so far, has
   anybody
   > > > seen any mention of any NSA programs designed to subvert
   compilers?
   > > >
   > > > Compilers seems like an extremely prime target for manipulation,
   but as
   > > > far as I am aware there hasn't been anything mentioned about this
   yet.
   > > > Has anybody here heard anything that I haven't?
   > >
   > > Given that compilers are both a fairly easy to attack and amazingly
   > > convenient target, it wouldn't surprise me if the NSA has subverted
   a
   > > few specific compilers that are in common use. An attack of this
   nature
   > > has been hypothised since the early to mid-1980's. They would have
   to be
   > > amazingly dense not to have at least considered it.
   > >
   > > On the flip side, the NSA likes to do things where it has the least
   > > opportunity to be caught. Compiler subversion, while not "easy" to
   catch
   > > by any means, might offer too big a risk of being caught for them
   to do
   > > it. Being that they have a multitude of weirdly named programs
   > > specifically set up to compromise software, the evidence would lean
   > > towards they haven't done it but I'm sure it was, at the very
   least,
   > > discussed.
   > >
   >
   >
   >
   > --
   > Kelly John Rose
   > Toronto, ON
   > Phone: [3]+1 647 638-4104
   > Twitter: @kjrose
   > Skype: [4]kjrose.pr
   > Gtalk: [5]iam@kjro.se
   > MSN: [6]msn@kjro.se
   >
   > Document contents are confidential between original recipients and
   sender.

References

   1. mailto:hozer@hozed.org
   2. mailto:cypherpunk@cpunk.us
   3. tel:+1 647 638-4104
   4. http://kjrose.pr/
   5. mailto:iam@kjro.se
   6. mailto:msn@kjro.se