Re: Opportunistic signatures - a proposed design

Atom 'Smasher' wrote:

> the hole is a spammer subscribing to the list, and sending out as much 
> spam as they can, before they get turned off. what's the solution for 
> this... spam filters on the lists? moderated lists? unmoderated posts 
> granted to users after they've "proven" their value to the list? new 
> members must be voted in?
> 
> collecting addresses and spoofing the list are not practical... on most 
> lists, a small percentage of the members do most/all of the posting. 
> this also takes more work than spammers are interested in.

fine.  This saves me work.  Let's just ignore mailing lists and let 
somebody else solve the problem.


> if someone can afford to send out enough email where that's a concern 
> (ebay? amazon? buy?), and they don't want to stamp it, there's a good 
> chance that they're a spammer.
> 
> OTOH, let's say i subscribe to a mailing list, 
> <[email protected]>. well, having subscribed to a mailing 
> list, i can accept stamps for that address.
> 
> seriously... this SHOULD be a burden to anyone doing a mass mailing! i 
> think of that as a FEATURE, not a BUG.

I understand.  But there are a serious number of "know it all" types who 
insist on getting their knickers in a twist over the mailing list issue. 
  We have to be prepared to answer and if the answer is going to be 
mailing lists can just go pound sand, then let's give them a bucket full 
and a hammer.

>> at the egress point for any aggregation of any number of e-mail users, 
>> traffic patterns start to look significantly like a spammer.  The 
>> greater the aggregation, the greater the likeness.
> 
> =============
> 
> if it were true that spamminess of an email can be correlated to the 
> volume of email passing through a node, then spam filters wouldn't work.

spam filters have many features they can measure by content, volume, 
addresses, domain name validity etc.  Hashcash is a rate limiter and 
primarily affects spam through the slowing down of how fast you can send 
it.  Now, look at what I said in that light.

> 
> 500 users minting stamps on 500 desktops is not a significant burden. 
> 500 users having all stamps generated by an outbound "stamp server" 
> could be.

yup, that's right.  Unless of course the organization says "we're not 
touching the desktop.  If you tell us to, go pound sand and have a nice 
day."  In my experience, this is any organization with more than 100 
desktops.

> thus, a market will exist for an outbound "stamp server", specially set 
> up for minting stamps. i suppose that's inevitable... but when they hit 
> the surplus market....

yes.  They certainly well and I've worked out how to take advantage of 
"spare CPU capacity" within an organization.  It might be a CPU cycles 
sponge as it were


> but then you're talking about domain signatures, not user signatures? 
> yahoo already invented that. and home users aren't going to have a 
> "secure" box anywhere... not as long as M$ is in business. so the 
> problem of stolen signing keys might be reduced for companies that can 
> afford to not keep keys on desktops, but the problem doesn't go away.

no, that is not what I am talking about.  I am talking about individual 
signatures associated with a given e-mail address.  I was talking also 
on the context of an enterprise or ISP environment.  The same technique 
will also scale down to the individual user.

Personal boxes will get more secure.  They are significantly more secure 
than they were five years ago.  Continuing to dither about stolen keys 
due to viruses and other intrusions will only serve to keep security off 
the desktop.  If there's nothing there to protect, why protect it.  As 
soon as there is something that to protect and a few people get burned, 
then they will be action on many fronts to make the desktop more secure 
and quickly.

If you're going to continue to dither about this, you might as well bend 
over, grab your ankles and say "I'm ready for my close-up Mr. intruder" 
because there isn't going to be a solution that is either practical or 
one that will make you happy.

> i would assume that a machine gets owned, and between 2-3am that night 
> the spam will start... send out a flood of spam while the mark is 
> sleeping, and when they wake up they'll have a lot of people ~very~ 
> angry at them.

see my point above.  If there's something to protect, the people will 
demand some way to protect it.  It's not going to happen a minute before.

>> If you take care and protect your keys with passphrases or external 
>> devices, you are much better shape unless you machine has been 
>> compromised at which point, you are screwed
> 
> ===============
> 
> for >99% of users, those precautions will not be taken.

at which point, you are screwed.  And then you will demand protection 
from it happening again.

> agreed then: M$ is the problem ;)

actually, computers or the problem.  It doesn't matter which OSes out 
there, compromises will happen, people will get burned, lawyers will get 
rich.

in all this argument of risk factor vs. risk factor, one thing neither 
of us can show is actuarial tables.  Until you can show that for a given 
number of hours on the net running particular piece of software etc. you 
have a 0.000x percent intrusion rate, all discussions of risk are just 
blowing smoke.  100 percent fine Caribbean smoke.

when risks are categorized, people will demand protection analogs of 
police, fire, insurance.

> assuming a compromised machine, the nice thing about hashcash vs signing 
> keys is that it throttles the rate at which spam can be sent...
> 
> compromised signing keys on a compromised computer would essentially 
> guarantee a DoS attack against everyone in the address book... the only 
> limiting factor is the speed of the connection.
> 
> a compromised machine in a hashcash universe would be limited to sending 
> out what... ~30 emails/sec? and that's at 100% load. my mom would notice 
> that, and call someone (not me, i dunno nuthin about windoze).
> 
> 
> summary: public keys can prevent forgery, not spam.

understood.  But the human factors aspect of dropping the cost of 
hashcash and substituting a signature as a forgery resistant channel 
between two parties, is hugely beneficial to the end-user.  The reason 
all security measures fail today is because of human factors.  Why do we 
still use passwords?  It's because they suck less than all the 
alternatives.  Why is e-mail plaintext?  Because it's easier to use than 
trying to remember a passphrase.  Why do e-mail and browser applications 
remember passwords for you?  Because it sucks less than remembering 
passwords.

so, we can go around this barn again or we can figure out a good, human 
factors friendly way of minimizing stamp load when it is really needed. 
  This is important because a large number of attacks against hashcash 
are political.  They are "I'm not touching my desktops", "why should I 
spend stamp time every time I send a message to someone I know","what 
about mailing lists", and a host of others.  They all appeared to boil 
down to a resentment to spending CPU cycles unnecessarily.  but what 
they really are is failures of human factors.  And when it comes to 
human factors vs. security, human factors always win unless there is 
some dude with an M-16 standing near you.

---eric