Hacker News Reader

Comment by rumcajz | Hacker News Reader

This is a procedural, not technical problem. It almost seems like the standardisation process open to everyone just enables everyone to insert their own backdoors into the standard.

One interesting way to solve the problem would be to allow differenct mutually hostile entities to define their own standards (US, Russia, China, FSF, Pirate Bay, whoever) and then encrypt using all of them.

That way, even if there is backdoor in each protocol, the only way to decrypt would be all those disparate players to cooperate. It would be like a vault with multiple keys possesed by different people.

View on HN

Thats an interesting concept, I wonder if Russia and China have their own standards/protocols to use?

I guess we could get the same effect by encrypting using 3DES, then AES, then blowfish, twofish and then RC4.

OpenSSL supports these Russian GOST standards:

* md_gost94 message digest algorithm

* gost89 symmetric encryption algorithm with 256 bit key

* gost94 public key algorithm with 1024 bit public key

* gost94cp public key algorithm with 1024 bit public key (CP mode1)

* gost2001 public key algorithm based on elliptic curves with 512 bit public key

* gost2001cp public key algorithm based on elliptic curves with 512 bit public key (CP mode1)

Should I make a simple script to use gnutls to encrypt a file with AES then openssl using gost89, then 3des ?

Why not? Seems like fun.

> I guess we could get the same effect by encrypting using 3DES, then AES, then blowfish, twofish and then RC4.

TrueCrypt supports cascaded encryption in XTS mode, see [0]. AES-Twofish-Serpent, combined with a decent password and Whirlpool hash algorithm (it is used for HMAC and for mixing the RNG) should be pretty secure, IMO. Anyhow, we don't have plenty of really good and checked publicly available alternatives for symmetric encryption.

[0] http://www.truecrypt.org/docs/cascades

Counterintuitively, sequentially applying additional crypto doesn't necessarily improve security; in theory, it may actually damage it.

Any pointer to explanation of the phenomenon?

If the keys used for each cipher are independent it should not weaken security. See section 15.7 of Applied Cryptography (http://www.cse.iitk.ac.in/users/anuag/crypto.pdf).

None of this helps you if the random number generated that created your key is compromised.

Even if you use different random number generators?

I suppose you could, yes. If you wanted to make all this happen algorithmically you'd end up with a super-sized key that was a concatenation of all the keys you generated for each of the sub-algorithms, where the sub-algorithms would each consume their respective key chunks.

You don't have to concatenate your keys, simply XORing the random bitstreams together works fine (my intuition says that XORing them would provide a stronger result).

  echo hello | crypt -usa | crypt -russia | crypt -china

Sounds ideal, until you realize that building and operating proper encryption is too hard for some of these organizations (PirateBay, FSF). For the rest, they are only mutually hostile on the surface. I am sure when it suits them they will get together and sell all our asses to each other.

You mean like Obama giving a tool to read his email to Putin? Unlikely IMO.

More generally, something along the lines of: "You give us Snowden/limit his asylum claims/etc etc, and we give you favourable trade terms".

Or maybe we work out a deal over Syria.

Seriously.

True, but Russia, China and the US all have a shared interest in working (together?) against Islamic rebels/separatists/militants/terrorists, so they may well cooperate more than you might at first suspect. If you could get some Islamic militants to (develop?) contribute a cryptosystem, then you would be more convincing.

One could argue that at least mass-surveilance would be much more difficult to pull off... But I am afraid that the governments would start cooperating (in 1984 style).

Iran should be able to do that.

Oh god, not again with the "war on terror".

It's arguable that their interest is to fight against "Islamic rebels/separatists/militants/terrorists".

One might say that they instead have a political and economical agenda disguised as war against terror.

If so, they would certainly have conflicts of interest, at least at some point.

I think something like W3C would be better.

W3C or IETF are open by their nature. Thus, there can be multiple backdoors inserted by multiple parties. So if you use multiple such standards to encrypr, it may happen that there's an entity that has backdoor in all of them.

A standard from NSA, on the other hand, is guaranteed to have exactly one backdoor. Same goes for KGB etc.

Reply on news.ycombinator.com