🚀 go-pugleaf

In <40nfll$tr@sundog.tiac.net> Jim Williams <williams@tiac.net> writes:

>billman@centcon.com (William Rini) wrote:
>>
>>We could go on forever talking about the ways that MS COULD get
>>information out of your computer, and no matter how much proof you
>>presented Mr. Williams  would never be satisfied, because in his mind
>>he has made a judgement, facts won't sway him from his version of
>>reality.  He calls it a court of public opinion and appoints himself
>>judge over what other people should think.  
>>
>
>Wrong.  All I require for proof is the sources, the build paramters, and
>the development platform.  (I do want to be able to build exactly the
>system Microsoft claims to have released and then do a byte compare of
>the results.)

There is a guy, called Ken Thompson, who brilliantly proved that this is
not enough.  He installed a trapdoor in the Unix login program without
touching the sources of the program, by hacking the C compiler.  Until
here, nothing special.  But he went one step further and hacked the C
compiler to recognize when it was compiling its own sources and add the
code needed to hack login.c automatically to the new compiler.  So, you
obtained a hacked C compiler starting from clean sources, by compiling
them with a hacked compiler.  Got it?  Since (almost) nobody is using
his own compiler, you got a hacked compiler and a hacked login program
starting from completely clean sources!  You have all the sources, yet the
trapdoor is there, installed behind your back.

I wouldn't expect Microsoft to be able to implement this technology into
their compilers, but it isn't impossible.  If you have to use a single
binary executable provided by someone else, you have to _blindly_ trust
that person (or to disassemble, with your own disassembler, of course :-)
the executable before running it. 

Dan
--
Dan Pop
CERN, CN Division
Email: Dan.Pop@mail.cern.ch 
Mail:  CERN - PPE, Bat. 31 R-004, CH-1211 Geneve 23, Switzerland