Hector Martin

Did Cure53 get arbitrary plaintext exfiltration across many clients? Serious question. I missed this result. Link?

"Arbitrary" is, er, rather generous. But sure, Cure53 didn't get the automated exfiltration, but did set the stage with the partial decryption story. My point is rather about the FUD that academic security teams seem to be rather prone to lately.

I would go further: the fact that you can maul encrypted PGP emails in various ways was known well before Cure53. It’s been known since the late 1990s. But nobody has ever taken it seriously enough to comprehensively address it across all clients. Guess why?

IMO, because active content in e-mails is an ancient horse that has been beaten to death and the fact that this is *still* a problem just demonstrates a pervasive failure of email clients to take privacy seriously, PGP completely aside.

This isn’t just about active content. It’s about the fact that active content can be subtly and comprehensively inserted into encrypted emails using very sophisticated cryptographic techniques, AND that content can be made to exfiltrate other encrypted portions of the data.

When you simplify it to “active content” you do a disservice to both the cryptographic sophistication of the attack AND the severity of the vuln (particularly for S/MIME). This is why things don’t ever get fixed.

I'm not considering S/MIME here at all since most of the FUD seems to be about PGP. I couldn't care less about S/MIME and have not analyzed that vector.

The PGP vector is bad enough.