Many of the emails released by the Justice Department from the Jeffrey Epstein investigation are full of garbled symbols such as:
Ciphertext is so ubiquitous that it has fueled conspiracy theories that it could be some kind of code. But as believable as it might be for a cabal of elite sex-traffickers to communicate in a secret language, the reality is probably more boring: The symbols are likely artifacts from the way the Justice Department converted emails into PDFs.
“The glyphs and symbols are probably some artifact of a bad conversion process,” said Chris Prom, a professor and archivist at the University of Illinois Urbana-Champaign. Specifically, the symbols look like remnants of Multipurpose Internet Mail Extensions, or MIME, a 30-year-old standard for encoding e-mail. The protocol that underlies email transmits messages as short strings of simple ASCII characters, so when people started writing longer messages and tried to include formatting and symbols, MIME was developed as a way to encode them in ASCII.
In MIME, “=” is used to signal either that a string of text should be broken and rejoined for transmission—a “soft line break”—or, when followed by two other characters, that it should be converted to a specific non-ASCII tag. For example, if you wanted to actually write “=” in an email, it would be encoded as “=3D”. In normal use, the recipient’s email client decodes these symbols before displaying the formatted message.
Regardless of the software the Justice Department used to extract the emails and convert them to PDF, some of the decoding appears to have messed up, said Peter Wyatt, chief technology officer of the PDF Association, who examined the batch of Epstein documents.
“It was in the news and there was a lot of PDF,” he said. The association conducted similar analyzes of the Mueller report and the Manafort documents. “Generally speaking, we’re interested in anything related to PDF. It’s something we do and we’re good at.”
The clarity of the text and the URL led Wyatt to believe that these documents were digitally extracted and then converted to PDF, rather than physically printed and scanned as was the case in the Mueller report. “So things have gotten better since then,” Wyatt said.
Specifically, the DOJ likely extracted email data, converted it to PDF, and then redacted it. To remove metadata from the document and bake in the redaction so that the black bars cannot be removed, they convert the documents to image files such as JPEG before conversion. rear to PDF. The software used to initially extract and convert the data also captured parts of the underlying MIME format instead of decoding it correctly. Or more simply: emails, sometimes partially decoded, converted to PDF, converted to JPEG, converted to PDF.
At least that explains the abundance of “=”. But it doesn’t fully explain why “=” sometimes replaces letters, like the “J” in “Jeffrey”. No one I spoke to could definitively answer that question, except to say that email is difficult and converting it to PDF is harder, and the DOJ was converting a lot of documents in a hurry. (Introductory notes were also significantly inconsistent in the files.)
Prom thought it might be a character set conversion problem, which he often saw when the archive tool he tested couldn’t find a particular character set or font that the email server was using.
Craig Ball, a forensic expert who teaches at the University of Texas at Austin School of Law, pointed out that different email clients implement the standards in slightly different ways, adding to the difficulty of conversion. “My hunch is that this is an incompatibility between the code pages used by the sending mail client (possibly BlackBerry) and the application used to print the messages to PDF,” Ball wrote. “The presence of BlackBerry and iPhone signatures in these emails suggests that the messages passed through multiple systems with different encoding procedures, complicating decoding issues during PDF generation.”
“You’re looking at hundreds of different methods of converting these files from hundreds of different people using whatever software they had available, some of which may have been good, some of which may not have been good,” Prom said.
“The PDF standard is quite complex,” wrote Prom. “And email to PDF is particularly challenging.”