Subj : Re: ChatGPT to obfuscate on the Dark Web
To   : paulie420
From : poindexter FORTRAN
Date : Thu May 04 2023 07:23:00

-=> paulie420 wrote to All <=-

 pa> Oxymonster, a darkweb drug dealer, was arrested while he was attempting
 pa> to attend the beard and moustache championship in Austin, TX. LEO knew
 pa> he was shutting down the darkweb store for a particular 4-day period,
 pa> as he announced it on the darkweb markets. They way the LEOs put this
 pa> together was in conjuction of knowing the dates and scraping other
 pa> social media's for posters announcing 'vacations' of the same time
 pa> period AND by linking his grammatical way of writing.


 There was a twitter bot called something like @pleaserobme that looked
 for people talking about vacations online and posted times they'd be on
 vacation along with meatspace coordinates gathered through OSINT.


 pa> I attempted to use ChatGPT to work around this by creating a prompt
 pa> where AI would rewrite my message posts in a random style - or, a
 pa> prompted style. It worked out really well, and with several tries I was
 pa> able to re-prompt ChatGPT to even m0re intelligently 'randomize' the
 pa> messages. This could be valuable for anyone who needs to not be tracked
 pa> from the way they write.

 I've been watching YouTube videos about using OpenAI, and one of the
 interesting bits was tuning the style of text. "Make me sound younger,
 more casual" or "Make this legal agreement sound less formal" is an
 interesting twist - instead of using AI to create the content, use it
 to change the look/feel of the content.

 The concept of "grammar fingerprinting" is certainly something to be
 aware of, and ChatGPT is an interesting wrinkle. Let's see if law
 enforcement tries to paint AI as a tool for illegal activities.



.... RAW DATA FOR RAW NERVES
--- MultiMail/Win v0.52
 * Origin: realitycheckBBS.org -- information is power. (700:100/20)

.