Re: [Lit.] Buffer overruns
From: infobahn (infobahn_at_btinternet.com)
Date: 01/14/05
- Next message: Marcel Martin: "Re: WAS Frobenius, so good?"
- Previous message: Paul Rubin: "Re: OTR - Truly private conversations"
- In reply to: Bryan Olson: "Re: [Lit.] Buffer overruns"
- Next in thread: David Wagner: "Re: [Lit.] Buffer overruns"
- Reply: David Wagner: "Re: [Lit.] Buffer overruns"
- Reply: Bryan Olson: "Re: [Lit.] Buffer overruns"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Fri, 14 Jan 2005 15:55:03 +0000 (UTC)
Bryan Olson wrote:
>
> infobahn wrote:
> > Bryan Olson wrote:
> >
> >>infobahn wrote:
> >>
> >>>And FWIW, I agree entirely. Get the spec right, get the program
> >>>right, and wave bye-bye to buffer attacks.
> >>
> >>How do you know when to give the wave?
> >
> > At about the same time that you gain *complete* assurance that
> > automatic bounds-checking (ABC hereafter) doesn't leave you open
> > to buffer attacks.
> >
> > ABC adds overhead, so it had better be able to pay its way. So
> > we'd better be SURE that it works. But how can we be? Even if you
> > validate your ABC compiler's source code, you have to use a
> > compiler to compile it.
>
> Where have you been for the last few decades?
Programming. Well, for some of the last few decades, anyway.
> In the ocean of
> exploits, there've been plenty of naive errors, and plenty of
> subtle errors. Compiler and run-time-system bugs comprise
> barely a drop,
...as far as we know; but, without spending a fortune finding out,
we can't know for sure.
> with the exception of systems intended to safely
> run code that may be written by adversaries. We know how to
> make automated methods vastly more reliable than manual
> checking.
Presumably you make them more reliable by checking them - manually!
>
> A while ago you advocated code re-use. Sophisticated compilers
> and run-time systems *are* code re-use.
Sure. But if you haven't got the source, you can't be sure. And
even if you /have/ got the source, you /still/ can't be sure.
> [...]
> > Absolute security is impossible if you also want to get anything
> > done. That doesn't mean we shouldn't try for security, of course.
> > But it's a fair indication that we should do the best job we can
> > with the best tools available, and what "best" means depends on
> > circumstances.
>
> Wake up. Check the circumstances. Sure there are still cases
> to use C or assembly. For most implementation, we now have
> better tools. Unfortunately we also have antique programmers
> who think choice of tools hardly even matters to security.
I, on the other hand, am still very young. Hardly even vintage,
let alone antique. And I think the choice of tools is extremely
important. But I'm not prepared to put down my chisel just because
some damn fool once cut off his own thumb.
- Next message: Marcel Martin: "Re: WAS Frobenius, so good?"
- Previous message: Paul Rubin: "Re: OTR - Truly private conversations"
- In reply to: Bryan Olson: "Re: [Lit.] Buffer overruns"
- Next in thread: David Wagner: "Re: [Lit.] Buffer overruns"
- Reply: David Wagner: "Re: [Lit.] Buffer overruns"
- Reply: Bryan Olson: "Re: [Lit.] Buffer overruns"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
