Re: Re: Fuzzing project brainstorming

[Daniel Kahn Gillmor <dkg () fifthhorseman net>]

On 11/20/2014 02:23 PM, Hanno Böck wrote:
> On Thu, 20 Nov 2014 08:52:15 -0800
> "M.T. Roebuck" <marvint.roebuck () inbox lv> wrote:
>
> > Maybe my problem is that your proposal seems herculean to me but
> > can't help to think it's a reminder or sign that we need to think
> > past the current state of things.
>
> Compared to "starting from scratch" starting such a fuzzing project is
> not herculean, it's more like grabbing the low hanging fruit.
>
> But arguments alike come up every now and then. Basically you'll hear
> two things: "We have to mitigate / sandbox" and "please rewrite
> everything in [insert favorite non-C programming language]".
>
> I don't want to downplay either of these approaches. It's just that you
> have to be realistic. Nobody will rewrite everything from scratch in
> rust/go/haskell/whatever any time soon. There are a few interesting
> projects that try to rewrite key sofware in safer languages (mitls and
> servo come to mind), but they are few and none of them is in a
> production state.
>
> Our systems we have today - the ones we use to have this discussion,
> manage our bank accounts and surf the web - have imperfect software
> written mostly in unsafe languages. I feel fuzzing can improve the
> state of things a lot.

I agree with this sentiment.  I also think this is likely to be a
herculean effort, and hopefully not quite a sisyphean one (the boulder
should be able to move up the hill a little bit each time).

I'm really happy that you're pushing on this, Hanno.

even if the only thing that comes out of it is a classification of which
projects/libraries insist on "trusted input" that would be a very useful
outcome.

        --dkg

Attachment: signature.asc
Description: OpenPGP digital signature