Slashdot Mirror
Distrustful US Allies Force Spy Agency To Back Down In Encryption Fight (reuters.com)
schwit1 shares a report from Reuters: An international group of cryptography experts has forced the U.S. National Security Agency to back down over two data encryption techniques it wanted set as global industry standards, reflecting deep mistrust among close U.S. allies. In interviews and emails seen by Reuters, academic and industry experts from countries including Germany, Japan and Israel worried that the U.S. electronic spy agency was pushing the new techniques not because they were good encryption tools, but because it knew how to break them. The NSA has now agreed to drop all but the most powerful versions of the techniques -- those least likely to be vulnerable to hacks -- to address the concerns.
" In interviews and emails seen by Reuters, academic and industry experts from countries including Germany, Japan and Israel worried that the U.S. electronic spy agency was pushing the new techniques not because they were good encryption tools, but because it knew how to break them."
The NSA is widely believed to have done exactly this when it recommended particular elliptic curve constants quite a few years back.
Once you've betrayed people's trust, you're going to have a hard time convincing them you're worth trusting with anything that matters ever again.
#DeleteChrome
New "ciphers".
Specifically, two new families of block ciphers called SIMON and SPECK. These ciphers are designed to be extremely fast, which is good because although AES is fairly fast on "big" hardware" or on large quantities of data, it can be a bit sluggish when used in extremely constrained environments on small amounts of data. In particular, its key schedule its heavy, so changing keys is slow. SIMON has been designed to make it particularly cheap in purpose-built hardware while SPECK is designed for very fast software implementations. Both are very, very fast on both hardware and software, though. The 128-bit version (block size and key size) of SPECK, for example, encrypts at about 1.25 cycles per byte on an i5 on long messages, and is almost as good on short messages. That's crazy fast.
Academic cryptanalysis of the ciphers has so far shown them to be quite solid, with a very good margin of security (meaning that cryptanalysts have only been able to break significantly cut-down versions of the ciphers, quite far from full versions).
Same trick.
Possible, but doubtful. In fact, the experience with Dual EC DRBG actually makes it significantly less likely, IMO. They tried to pull the trick with that, but it didn't work because academics discovered the mathematical structure that made the backdoor possible. That has to make them worried that the same thing would happen again, and in fact the trick would be much harder to pull off with symmetric block ciphers. The thing about elliptic curves is that they have rich mathematical structure which can be exploited in clever ways (this is what makes them useful for public key cryptography) by choosing the right curves. But symmetric key block ciphers like SIMON and SPECK don't have that, making it much harder to design back doors in.
It's not impossible that the NSA has some technique that can break these ciphers -- which are actually quite similar to ciphers produced by public cipher designers -- but it really seems unlikely. Nevertheless, once burned twice shy. I don't blame standards bodies for being reluctant and waiting for public cipher designers to produce algorithms with the desirable properties of SIMON and SPECK, but without the concern about origin.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
The key doesn't need to be the same length as the cleartext, it can be considerably shorter. This does weaken the encoding, but not fatally. You just need to encrypt the message before you encode it with the one-time pad with a code that's difficult to recognize. The more you shorten the key, the weaker the encoding, but shortening it by 50% is still quite safe if you use a decent encryption of the cargo.
Perfection isn't impossible, but is hideously expensive.
That said, any code that depends on factoring large primes is weak when used against quantum computers. And they may not be here today, but I wouldn't make strong bets about next year in secret government offices. So if it's worth it to you, by all means use one-time pads. And most of the expense of using them is in the transmission of the info, so you might as well use the most secure version. You can get a pretty good set of random numbers by processing a web cam of a candle flame, but turning that into terabytes of good random numbers could take awhile.
I think we've pushed this "anyone can grow up to be president" thing too far.
The "merits of the algorithm" is communally undefined if the design party is keeping secret the existence of differential cryptography—or any other advanced mode of attack—as IBM and the NSA once did with the DES. It was pretty clear that something fishy had gone into the design of the S-boxes. Whether fair or foul is impossible to decide when you're on the outside looking in (turns out, for DES, it was fair—foul play was confined to mandating a short key length).
What people don't understand is that as much as the Americans would like to read everyone else's traffic, it's far worse if any backdoor leaked to an adversary (your whole financial system is protected by these codes), so they were sensibly reluctant to put one in—until they invented the one-way back door, where only the designers could ever know. Unable to resist the siren call of this new brass ring, the NSA immediately blew their entire history of trust (which had always been more out of enlightened self-interest than gentlemanly) into a giant mushroom cloud.
It remains difficult to decide whether "merit" can be debated in these matters on a level playing field.
On the other side of the coin, while I'm far from a serious cryptographer, Specks' ARX design does not appear to leave many places for newly discovered snookery to hide itself.
That said, banning the runt versions smells like prudence to me, as any covert American attack is probably a combination of a downgrade attack—tricking a cipher to operate at less than full strength (world and dog are not freaking out over the Intel Management Engine for no reason)—perhaps injecting some known plaintext, finished off with a giant can of precomputed whup ass (the mechanism of attack one can best keep confined to your side of the fight is a multimodal attack).
Once you take the downgrade attacks off the table, it's a lot easier to swallow the inequitable debate on merit as a pure cipher.
Not buying it. I really don't see how you performed that neat dissection of history from technology from capabilities, without the use of a white glove and a black hat.
____
Addendum:
Researchers Find a Way to Disable Much-Hated Intel ME Component Courtesy of the NSA — 28 August 2017
True to form, the NSA's greatest terror is being hoist by their own petard.
They don't advertise this fear, because they prefer to viewed through the do-unto-others side of the lens. Trying to turn these weapons into technological diodes is an enormous practical constraint.
That, and resource saturation (what they can do and what they can afford to do are two different beasts) are in my experience the only reliable external vantage points for 99.999999% of the planet's population incapable of wading into the merit debate at anywhere near eye level.
Your sig is a lie!!! ;)
I do not want your cheap brainburning drugs. They are useless for work. And I am a working man today.