[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Is Graham-John's inane spam robogenerated?



Dale Thorn wrote:
> Igor Chudov @ home wrote:
> > Dale Thorn wrote:
> > > Igor Chudov @ home wrote:
> > > > stop spamming the list thorn
> > >   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > > What the hell is this?
> 
> > That's an illustration that it is easy to post things that look
> > autogenerated, but are not. On the other hand, it is possible to
> > autogenerate things that look non-trivial to a novice.
> 
> I'll agree readily with the latter sentence, but not with the
> former, if a fairly large number of variations are involved.
> After all, who would bother with that much precision typing?

I am not so sure that it really was precision typing. (and would like to
look at evidence)
 
> > One gentleman from a third-rate educational institution is known
> > for sending tons of lisp-generated articles to one of the moderated
> > newsgroups, just to annoy moderators. They do look like they are created
> > by someone with a rudiment of inteligence.
> 
> Yeah, that's the goal of the spambots. They're actually useful tools
> for combatting the elitist parasites, er, tenured professors who
> troll these net forums so much. Problem is, the sheeple get confused
> about who's doing what to whom...
> 

The paragraph above reminds me of The Right Reverend Colin James III.

He was also trying to combat elitist professors. That is a long and
happy story.

I think that a perfect spambot is possible and is a great exercise in
programming. It is also a cool and very creative idea, and as someone
suggested earlier, it can be created on the basis of cbcb. The net
result of the spambot would probably be a huge scandal and lots of
people leaving usenet. Some of them would be tenured professors. I see
few people who would benefit from it though.

Along the lines of poetry festivals and spambots, I may suggest this.
When I was 16, I wrote a prose writing program in Pascal. It read a long
text and created a table: as the key, it had pairs of words, and as the
data, it had list of all words that follow the pair in the index. The
table was generated by a single pass through the source text, where
there was a moving 3-word window and first two words were used as the
key to the third word. The window moves one word at a time.

The program then attempted to generate intelligent-sounding garbage, in
the following way. It started with a random pair of words from the
source text. It then looked up the table and selected(**) the word that
was most frequently used after these two. Then a moving window moved one
word right to the next and took the last word (which was just selected)
and the word before last as the key into the table, and did that ad 
infinitum.

The loop repeats indefinitely. The text that results looks like it was
written by a schizophrenic -- it is more or less correct grammatically,
uses more or less compatible words and seems to make sense, but the
meaning seems to evade the reader. It is an extremely strange and annoying
feeling.

(**) The problem with this algorithm is that after a while, it starts
looping.  To fix that, the process of selection needs to be randomized
somewhat. The possibe randomizations are obvious.

To apply this to poetry and following-up spambots, it can do the
following [besides forging headers, etc]: for each message, read it,
create the table, and follow up with "I agree" and a schizophrenized
version of the quoted article.

It can also use USENET as a bigger source of the triples. I strongly
suggest to build one table per newsgroup and not mix diff. newsgroups
together.  This way, spambot posting to comp.lang.eiffel would talk
about Eiffel and contravariance, and a spambot posting to
soc.culture.russian would talk about lying homosexual purebred sovok
forgers.

If we think about it for long enough time, this algorithm guarantees
that spambot-generated messages will always be on topic in the
newsgroups that are being spammed. That is going to perplex people 
very much.

	- Igor.


``In my final assault to save time for all men to have Eternal Life, I had
to face eons of time limits (negative micro-split second, split second,
etc., time limits) since I was born at Hanceville and could only make
ten mistakes in one locality or else it would have been over for all men
in Eternity as they would have been exterminated in the spirit and dead
forever in a lethal deadly proton.''

``u.s. atty d.blair watson returned my call today,thank you, he
received a letter from ok. atty general office referring my info
to him about gardner ks 8-10-95. i talked to watson today 3-5-97
watson told me that its not against the law for a federal 
operative to intimadate a person from entering the u.s.federal
courts building, i find that hard to belive but thats what he
said, i guess what one person considers intimadation a other
person might not,i told him about the part of the setup were it
appeared a person was going for a gun,when joe t*** pulled 
towards the FLAGED car and about hit it,its like well did you 
see a gun? no....was i supposed to? would it matter if i had?
what if they shot at me and missed? i think watson missing the
point, if watson knows why          congress voted to stop
wiretap authourity from expanding,he would check out joe T***''