> Presenting text for the user to input via an image is one way.  The image
> can have a lot of noise to make it virtually imposable for OCR software
> to recognize without errors.  I am sure there are other ways.

But who generates the text? A malicous node could simply not generate any
text and 'pretend' they have. Unless, random peer(s) generate parts of the
key, and encode them as OCR-proof images which the user must enter.

> Although it will be a pain for the user.  How about having multiple
> (random) nodes verify the data with each of them appending the signature
> to the information.  Since replying to email is not a lot of trouble you
> could require say a dozen nodes send email to the user and verify that the
> email address is indeed valid.  It would be difficult for all of them to
> be cancel nodes.  Heck you can even make it larger then that.  Replying to
> all those emails will be a pain but it will only have to be done once.

This would work to ensure that only people that actually have access to an
email address can claim a key from it, but it still won't get rid of the
man-in-the-middle problem, simply shift it to the mailbox. However, it would
certainly up the difficulty of such an attack.

> If cancel nodes can be identified some way and one of the nodes
> verifying the information is a cancel node the information should be
> thrown out.
>
> Then there is the actual security of the email address is question.  But
> since the mail will be coming from so many different directions an
> attacker would really have to compromise the mail server....

Yes, I think some form of bad node detection should definitely be built into
such a system, like my suggestion of keeping 'routing slips' with each
message to find common nodes in paths that result in invalid keys.
Unfortunately, this destroys anonymity.

> > for example, what do you do if you use the network to look up the pubkey
of
> > someone you want to contact, and it turns up more than one key, at least
one
> > of which could be the key of an intruder trying to listen to your
> > conversation?
>
> In that case the information should be thrown out.

But what if the user has at some time created a second key under the same
email?

> I think there would have to be some way to make nodes accountable for
> there actions and be able to detect the cancel nodes.  I believe GNet is
> working on such a problem.

Certainly. (See my comment above).
I'll take a look at gnet :)

Nick


_______________________________________________
freenet-tech mailing list
[EMAIL PROTECTED]
http://lists.freenetproject.org/mailman/listinfo/tech

Reply via email to