Re: anon.penet.fi hacking

On Feb 23, Hal said:

    Unless Eli's suggestion works - having our remailers put out a
    random "From:"  line (perhaps just on mail to Penet?)  might cause
    Penet to issue a new pseudonym for that apparent new user.  This
    would be kind of wasteful from Penet's perspective - all those
    pseudonyms are never going to be re-used.  But it might allow this
    form of chaining, without compromising the pseudonym of the
    remailer operator.

I would be cautious about a random "From:" line.  I think penet will
probably reject input that at least has does not have a valid (but not
necessarily truthful) return address.

For a while, Miron Cuperman's wimsey remailer was generating a bogus
From address, something like "yeltsy@kremlin.vax.ru".  I tried
chaining this to penet to post to newsgroups, but my anonymous
messages never appeared in the newsgroups.  This was because,
I think, penet sends a confirmation back to the sender. Since
"kremlin.vax" is not in penet's net tables, this would cause
the confirmation send to fail; my hypothesis is that this also
causes the newsgroup post at penet to fail.

One good From address to generate, at least for wimsey, would be

        pool0@extropia.wimsey.com

but you would need to add some boilerplate explaining to the
recipient that "pool0" is a group address and a return message may
be seen by up to (number? 10-100?) persons unless it is encrypted.

Wimsey could also establish its own penet password and automatically
insert it whenever it detected a "to" address ending in penet.fi.

It could also substitute the penet anonxxxx address for "pool0" in the
above boilerplate. This would allow a penet return to pool0. Note that
I don't mind too much if people know that they can reach Edgar Swank
through pool0, because pool0 is also an address for many other people,
so I can "plausibly deny" that any particular message addressed to
pool0 is for me.

I'd like to point out that so far the wimsey remailer is the only
useful remailer from my point of view because it's the only one
which allows me to delete the automatic sig.  This is because it
only forwards encrypted text and discards any appended unencrypted
text.

The other remailers were supposed to add a "kill line" to do the
same thing, but as far as I know this never happened.

--
edgar@spectrx.saigon.com (Edgar W. Swank)
SPECTROX SYSTEMS +1.408.252.1005  Silicon Valley, Ca

> I would be cautious about a random "From:" line.  I think penet will
> probably reject input that at least has does not have a valid (but not
> necessarily truthful) return address.

I have no way to check the validity of an address, unless it's
syntactically illegal.

> For a while, Miron Cuperman's wimsey remailer was generating a bogus
> >From address, something like "yeltsy@kremlin.vax.ru".  I tried
> chaining this to penet to post to newsgroups, but my anonymous
> messages never appeared in the newsgroups.  This was because,
> I think, penet sends a confirmation back to the sender. Since
> "kremlin.vax" is not in penet's net tables, this would cause
> the confirmation send to fail; my hypothesis is that this also
> causes the newsgroup post at penet to fail.

No, the posting must have failed fort some other reason. The problem is
that you never know why, as the error messages don't reach you...

> Wimsey could also establish its own penet password and automatically
> insert it whenever it detected a "to" address ending in penet.fi.

This could be one solution. But what do you do with bounces due to
some user error?

> I'd like to point out that so far the wimsey remailer is the only
> useful remailer from my point of view because it's the only one
> which allows me to delete the automatic sig.  This is because it
> only forwards encrypted text and discards any appended unencrypted
> text.

This will be solved using MIME. The .sig killer used at anon.penet.fi is
a pain in the rear...

	Julf

>The other remailers were supposed to add a "kill line" to do the
>same thing, but as far as I know this never happened.

Your call for this went unacknowledged but nevertheless listened to.
It was not until a week or two after the sig-kill stuff was over that
I came up with a solution.

The next revision of the remailer will have something like

	Body-Termination-Regex: <regex>

The first character in the body that matches the regex, and every
character after it, will be dropped.  This not only makes it a
one-liner in perl (!), but it means that the user can be as
arbitrarily complex in recognizing sig blocks as the are able.
Of course, we'll document the most common of these:

	Body-Termination-Regex: ^--$

For those of you who know nothing about regular expressions, this
recognizes a line containing two minus signs and nothing else.  If
your signature adder does it some other way, it's pretty much
automatically supported.  You could also put more of your signature in
the regex to ensure that it doesn't interfere unexpectedly with body
content.

Summary: user-defined, almost every case handled, not automatic.

I hate my sample header field name.  Please, someone think up a better
one.

Eric

> The first character in the body that matches the regex, and every
> character after it, will be dropped.  This not only makes it a
> one-liner in perl (!), but it means that the user can be as
> arbitrarily complex in recognizing sig blocks as the are able.
> Of course, we'll document the most common of these:
> 
> 	Body-Termination-Regex: ^--$

Ok... Now.. Show me the regex that strips anything starting with '--',
such as '-----------------------------', except a PGP boundary line...

	Julf

>> 	Body-Termination-Regex: ^--$

Julf writes:

>Ok... Now.. Show me the regex that strips anything starting with '--',
>such as '-----------------------------', except a PGP boundary line...

OK.  It's a mess.  The backslash means line continuance.  Remember that
concatenation binds higher that alternation (|).

^--...([^B]|B[^E]|BE[^G]|BEG[^I]|BEGI[^N]|BEGIN[^ ]|BEGIN [^P]|BEGIN P[^G]\
|BEGIN PG[^P])

Eric

> >Ok... Now.. Show me the regex that strips anything starting with '--',
> >such as '-----------------------------', except a PGP boundary line...
> 
> OK.  It's a mess.  The backslash means line continuance.  Remember that
> concatenation binds higher that alternation (|).
> 
> ^--...([^B]|B[^E]|BE[^G]|BEG[^I]|BEGI[^N]|BEGIN[^ ]|BEGIN [^P]|BEGIN P[^G]\
> |BEGIN PG[^P])

Hats off to you, Eric! Much better than I could do! But... It still
doesn't strip off something starting with only "--" on a line by itself...

	Julf

Julf challenged:
>Ok... Now.. Show me the regex that strips anything starting with '--',
>such as '-----------------------------', except a PGP boundary line...

I posted something which didn't quite work, as Julf says:

>But... It still
>doesn't strip off something starting with only "--" on a line by itself...

^--(|.|..|...(|[^B]|B[^E]|BE[^G]|BEG[^I]|BEGI[^N]|BEGIN[^ ]|BEGIN [^P]\
|BEGIN P[^G]|BEGIN PG[^P]))

Some implementations don't support empty alternation, so that could be
changed with the ? syntax, since (|a) and (a?) are the same.

That should do it.

Eric

> ^--(|.|..|...(|[^B]|B[^E]|BE[^G]|BEG[^I]|BEGI[^N]|BEGIN[^ ]|BEGIN [^P]\
> |BEGIN P[^G]|BEGIN PG[^P]))
> 
> Some implementations don't support empty alternation, so that could be
> changed with the ? syntax, since (|a) and (a?) are the same.

Uh... Eric? Would it be ok to use it as a shocker example on the basic UNIX 
courses I do?

	Julf

Body-Termination-Regex: <regex>

	 The first character in the body that matches the regex, and every

BTW I think this is just the right thing.  Excellent idea!

dean

Body-Termination-Regex: ^--$

	 I hate my sample header field name.  Please, someone think up a better
	 one.

Alternatives:

Sig-Pattern
Sig-Regex
End-Pattern

dean