[hashcash] Re: considering hashcash
- From: "Eric S. Johansson" <esj@xxxxxxxxxx>
- To: hashcash@xxxxxxxxxxxxx
- Date: Tue, 24 Aug 2004 16:09:19 -0400
Atom 'Smasher' wrote:
i'm not sure i fully understand what you're describing here... it sounds
similar to domain keys, but per user, instead of domain... outbound
mails are automagically "signed" and a recipients MTA/MUA/filter stores
valid public keys...?
that's basically it. The only other detail is the introduction process
which is if I see a valid stamp and a key some number of times and I
have replied to that address, I will accept a signature based on that
key as a substitute for a stamp.
that could lend itself to a nice PKI model, where
one can either be included in a strong set only after N people vouch for
them (aka, sign their key), or anyone's key can be put on a "keyserver"
and keys with a certain level of trust can be accepted by recipients
(and of course keys and signatures can be revoked if a computer gets
0wn3d)... it really doesn't need to be pgp-level security, as much as a
significant nuisance for spammers to forge and/or replay.
the only thing I would quibble with it is the use of a central authority
in any of this. After all, one of the goals I've had for
camram/hashcash is complete off line stamping, signature and verification.
anyway, in a PKI model like that, anyone who signs a spammer's key would
quickly find themselves with no valid signatures on their own key, and
they'd be quickly removed from the WoT. i'm sure an interface for that
type of app (if/when it exists) would include an opportunity to revoke
all signatures of anyone who signed a spammers key.
now you have just demonstrated a potential ability ability to censor
communications. This is not good. Granted, one could just fall back on
generating stamps and but I am twitchy sensitive to this issue.
attached are some of my thoughts on domain keys... i'd have to think
about how much of that would apply to a user-centric key system, with
the assumption that users are not properly protecting their signing
keys... but, assuming that people would leave their keys unprotected on
vulnerable machines, then a "proof of work" model seems superior.
you'll never get people to "properly protect" keys. Doing so increases
liability to the individual especially in the face of unreliable system
software. Unprotected keys increases plausible deniability (a virus did
it, your honor) and I think that will increase acceptability for this
limited domain. You'll notice that I almost never use PGP even though I
have keys on the key server network.[1]
the two reasons for going to the key system instead of proof of work
systems is to give us a mechanism which will allow us to increase stamp
size to really hammer spammers without hurting ordinary users, as well
as giving a mechanism for mailing lists to be able to deliver traffic to
all users quickly and efficiently.
so, I personally could live with stamps all the time for everything but
mailing lists and big ones at that because I do all stand generation and
background and I just ignore it. The thing to remember about doing
stamps all the time is that it must happen at the end user machine and
not on the mail server because there are more resources available for
the stamp generation process (i.e. a benign variant of the zombie problem)
---eric
[1] that actually comes for a couple of reasons. Usability is the
first. When your hands don't work right, typing in passphrases on a
regular basis just sucks. Therefore, I would need to use stored
passphrases which puts me almost at the same security level as keys
without passphrases.
the second reason is that I can't remember passphrases without writing
them down. Therefore, it's another vulnerability based in usability.
what I want is the equivalent of my car keys. Say, on the USB bus. I
would argue for a modification of the USB flash memory system with a
light and a button. If you perform an operation requiring access to
your key, the light lights up, you push the button and access is granted
until the light goes out. Needless to say, the light goes out
immediately after pressing the button from the human's perspective.
It's only a rough idea but that's basically the kind of user interface
needed to make keys more secure with ordinary people.
--
Speech recognition in use. It makes mistakes, I correct most
Other related posts:
