Hacker News Reader

Comment by cronix | Hacker News Reader

For the same reason why the NSA siphons up all data from the net, encrypted or not, and stores it. They know one day they'll be able to crack todays crypt. Being able to decrypt a backup 5 or 10 years from now will still provide a lot of useful data, whether it came from the darknet or someones icloud backup while it was in transit to Apples servers. Encryption really only protects you from "today", not what will happen to it in the future. It's one reason I kind of chuckle to myself when people put such high regard into companies that proclaim to be more self conscious about protecting user data and what they do with it. It doesn't really matter if the data is still being collected and stored for later decryption. It's basically just a "delay" of when it can be read. The only thing stopping them, currently, is not having enough qbits in a quantum computer. Things are going to get, strange. https://www.youtube.com/watch?v=vNV_3PkA9WM

View on HN

Quantum cryptanalysis is no help with symmetric ciphers. All the data encrypted "at rest" via symmetric ciphers, which is what companies are usually bragging about, isn't affected.

Now, I'm sure there's a ton of data that the NSA will be able to break with quantum computing, because people often do transmit data by encrypting it with an asymmetric cipher, e.g. PGP. But... there's already work on quantum-proof asymmetric ciphers.

And TLS uses RSA, so maybe they can crack all the https traffic? Well, assuming today's algorithms are configured correctly, they employ perfect forward secrecy. That means the key exchange algorithms never transmit or store the encryption key, rather an emphemeral encryption key is recreated by both sides. Of course, even with PFS they do get all the metadata and certs when they crack RSA.

Quantum cryptanalysis is no help. It doesn't exist, probably never will and people should stop pretending like it is inevitable to crank their citation count up.

Just xor bit by bit against real random noise,embed it in a massive stream of other pure random, remember the offset, if it's really that important..

Don't roll your own crypto. If that was much better encryption, everyone would be using it.

I think we all agree rolling your own crypto is dangerous, but what siliconunit described is just a one time pad. Assuming your key data is truly random and unknown to your attacker, isn't this kind of the gold standard for uncrackable cyphered communication?

In theory, but getting that stream of "real random" and above all distributing that one time pad securely to the recipient (chicken, meet egg) so they can read the message make this highly impractical.

dont need the random really. just need an offline method for exchanging the pad :D or just use an unexpected channel like radio or so to send it

If interpreted that way, storing the noise and memorizing the offset, it amounts to having a privately stored one-time pad to be used as key... with a passphrase with as many bits in it as the offset. That’s probably not a lot of bits. You are better off storing the data, encrypted with a real pass phrase, wherever you would have stored the random noise stream that is needed to read the data in any case. Don’t roll your own crypto.

Using a one time pad isn't "rolling your own crypto".

It is also provably secure unlike every other cryptosystem.

Sure, except now you have an equal amount of data that needs to be stored somewhere. The pad is as long as the data. Either you have to encrypt that using some other means, or you do the weak "store offset" method.

I agree with not rolling your own crypto for your primary mechanism, layering it (in the correct order!) is guaranteed to be at least as secure as your strongest crypto.

That's what google did with CECPQ1, they use "new hope" which is a quantum resistant algorithm with traditional methods (X25519). That way if new hope is cracked, they are still using an industry standard you'd have to crack as well.

the only proven secure crypto is the one-time pad. rest is all junk relying on computation times to stay 'secure'. and this onee time pad you can never send over a wire to anyone because of chicken-egg problem. so this would be something you would need to exchange offline for decryption to be possible.

if you are not doing that, best stick with known good encryption schemes.

that being said you can implement a lot of encoding schemes like you say in ways that make it arbirarily hard for people to decide what is junk and what is data, to make it nearly impossible to crack especially if you say do that xor with random values etc. because if you'd receive a data intercept you have a hard time to rebuild the data from the junk and then decode it

Reply on news.ycombinator.com