On Wed, 2008-01-23 at 08:02 -0600, Les Mikesell wrote: > Matthew Saltzman wrote: > > >>> But the NSA would be at least as capable of introducing a hack that you > >>> could examine but not see as Ken Thompson: > >>> http://www.everything2.com/index.pl?node=Reflections%20On%20Trusting%20Trust > >>> > >>> I'd expect them to even be able to conspire with the CPU vendors to have > >>> certain undocumented opcode sequences do magical things. > >> Sure. You can believe whatever you want to. I am merely stating a fact > >> that the bar to do this with open source software is way higher than > >> proprietary software and in fact is the highest that anyone can > >> practically go. > > > > Also, in order to carry out a hack like that, you have to infect the > > toolchain somewhere along the line, so that everyone building the code > > is doing so with infected compilers.. With open-source code and an > > open-source toolchain, that seems pretty unlikely. > > > > Or are you suggesting, Les, that everyone's copy of gcc is derived from > > one built by the NSA and smuggled into RMS's lab at some point in its > > early history? > > How many people have contributed code and how much do you know about > them or their motives? But a more likely target would be the CPU Rahul's point (as I take it) is that at least OSS code gets a fair amount of peer review by a wide variety of people who don't necessarily share the NSA's nefarious motives. Way more than can be expected from proprietary code. (Think Diebold...) My point is that infecting an open-source toolchain is much harder than infecting a proprietary one, for the same reason. I'll certainly acknowledge that there is no such thing as perfect security. > companies since there are only a couple that matter and this could make > the compiler portion pretty much invisible. Is that any more paranoid > than thinking the major communication companies all have government taps > for everything passing through or that cell phones are all rigged so the > government can locate and listen at any time? Probably not... > -- Matthew Saltzman Clemson University Math Sciences mjs AT clemson DOT edu http://www.math.clemson.edu/~mjs -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
