| HN Mirror

At the risk of re-igniting the flame war, I want to explain why one-time pads are not the solution, because you kept insisting that I rejected your proposal because I wanted to be loyal to the pro-HTTPS crowd, and not because one-time pads won't actually work.

If you find this patronizing, I apologize. I do not mean to be. But you admit that you're not into crypto, and so I'm assuming you don't understand why your proposal wouldn't work. At the very least, I don't know what you know and don't know.

So, as mentioned, supporting HTTPS on your site adds the problems of OpenSSL's large code base, a history of OpenSSL vulnerabilities (Heartbleed being very well known), and the problems of PKI. These are all well-known, and are deemed an acceptable risk because the risks of NOT using HTTP are much higher.

Now, one-time pads definitely have a lot of benefits. First, they're the only algorithm mathematically proven to be unbreakable when used properly. They are extremely simple to implement, and using them for encryption instead of having to choose between the dozens of algorithms that OpenSSL uses could reduce your encryption library's memory and storage footprint.

Of course, the big caveat here is that they are unbreakable when used properly. This means:

- A pad can never be used more than once -- If a pad is used twice, then an attacker that has sniffed two encrypted messages can derive the pad, and therefore decrypt all message encrypted with that pad. This derivation is quite trivial, especially for text data. (see https://crypto.stackexchange.com/questions/59/taking-advanta...)

- The pad must be at least as large as the payload -- If the pad is too short, you have to repeat the pad, causing the problem mentioned above.

- The transferring of the pad needs to be done over a secure channel -- You have to assume that an attacker is always listening and can see all traffic. Even if the server tells the client "Go get the pad from this URL", then the attacker will get that URL as well. Also, keep in mind that the pad needs to be at least the size of the data being sent, which means that if you have to fetch a pad, you're doubling all bandwidth requirements.

- The random number generator used to creating a pad must be cryptographically secure -- Being able to predict the supposedly random numbers another system is generating is a pretty known attack and has had several proofs of concept. Generating the amount of cryptographically secure numbers to create a one-time pad large enough for transferring very large amounts of data just isn't feasible.

- One-time pads do not solve the problems of PKI -- One-time pads do not offer any method of authentication. They're nothing more than a form of symmetric encryption, and symmetric encryption does not attempt to solve the problem of authentication.

That's basically the gist of it. If you want me to explain anything better (Like what symmetric encryption is, and how it differs from asymmetric), I'd be more than happy to. I'm not a crypto expert by any means, but I like to think I have a strong grasp of the basic concepts and ideas.