Hacker News Reader

Comment by jayzee | Hacker News Reader

So here is an idea for an email feature/gmail plugin: 'Semantic Scramble' Basically, any sensitive email that you send to a bunch of people is automatically scrambled (retaining original meaning/correct grammar, just a small shuffling of words) so that each person gets a unique email. Easy to find out who leaked it. You could add a similar unique jitter for sensitive photos/images...

View on HN

Considering PG didn't strike down the story I presume to think he expected it to leak out or didn't ask it to remain confidential.

Re: "Semantic Scramble", this was/is used quite successfully in a university course I tutored. Students are supplied with Java code featuring a unique comment style (/ * */). So many students blindly copied code out of another student's assignment and submitted it as their own that we could effectively generate a map of the plagiarism without even looking at the code -_-

For images steganography would work easily enough and for "semantic scramble" see "Practical Linguistic Steganography using Lexical Substitution"[1] from EMNLP 2010 =]

[1]: http://www.cl.cam.ac.uk/~sc609/pubs/emnlp10frannie.pdf

Or pg realizes that once the cat is out of the bag, you can't put it back in. If he takes it down here, it will be posted elsewhere.

At least this way the cesspit that is Business Insider doesn't get any extra pageviews from YC.

But there's another idea: when you want to share a sensitive e-mail, you run it through a 'Semantic Scramble' so it can't be detected that you specifically leaked it.

Generally in high security environments, you don't transmit sensitive information on systems where you're one button (intentional or accidental) from sending the information to the public.

Clearly labeling every instance of the information with how sensitive it is also helps (per page or per paragraph).

It's unrealistic to build a separate semi-airgapped network for most business purposes, but email is broken for sensitive information -- some kind of document room/DLP/etc. system is probably a good idea.

That's called salting, and many firms already do it. I'm not aware of any software packages to automate this, but I haven't looked either.

The canary trap.

For those not familiar with the term:

http://en.wikipedia.org/wiki/Canary_trap

This is how the music industry does it with material they send out to different parties like reviewers or magazines with tools like Haulix, except they use watermarks that are undetectable to the ear but encoded in the file itself.

Wrt email however, I don't think Paul really minded that it was leaked or just realized it was futile to stop it, perhaps that's why he un-killed it.

PR people do this, it works

Wouldn't this be easy to defeat by changing words around yourself before distributing? It would either match to someone else's copy of the email, or not match anyone at all. It would need to be highly resilient to a lot of tampering, which doesn't seem reasonable for relatively short documents like emails.

Reply on news.ycombinator.com