crypto: privateEncrypt & publicDecrypt Implementation

[fuson]

Without this functionality i can't migrate from ursa module.

[calvinmetcalf]

I'm not sure the question, they are in iojs

[Fishrock123]

@calvinmetcalf Those are the opposite of what they are asking.

publicEncrypt & privateDecrypt vs (their) privateEncrypt & publicDecrypt, assuming they are wanting the correct things.

[fuson]

Fishrock123, correct.

[calvinmetcalf]

ah missred, if you turn off padding and use privateDecrypt on plain text you will get ciphertext that can be decrypted with publicEncrypt

[bnoordhuis]

@calvinmetcalf Is that process completely symmetric? My initial hunch was that the result from crypto.privateDecrypt() would get bumped to cipher block size even with RSA_NO_PADDING, and a quick test seems to confirm that.

[calvinmetcalf]

what do you mean by cipher block size in this context modulus size?

[bnoordhuis]

@calvinmetcalf This, basically:

var options = { key: key, padding: constants.RSA_NO_PADDING };
var encrypted = crypto.privateDecrypt(options, 'plaintext');  // encrypted.length == 128
var decrypted = crypto.publicEncrypt(options, encrypted);  // decrypted.length == 128
// decrypted consists of 119 zero bytes followed by the original plaintext

[calvinmetcalf]

yes it's left padded up to the modulus length, it likely does it when you do

var options = { key: key, padding: constants.RSA_NO_PADDING };
var encrypted = crypto.publicEncrypt(options, 'plaintext');  // encrypted.length == 128
var decrypted = crypto.privateDecrypt(options, encrypted); 

as well (can't check at the moment). It's enough functionality to write your own padding/unpadding on top of it

[calvinmetcalf]

to follow up , @bnoordhuis you are correct that it is not symmetric,

var encrypted = crypto.publicEncrypt(options, new Buffer('plaintext'));

throws an error unless you left pad it with zeros yourself, privateDecrypt left pads it for you because the output of publicEncrypt will omit a leading zeros (for instance with this key var encrypted = crypto.privateDecrypt(options, new Buffer('plaintext')); creates an output of 127 bytes.

[indutny]

It can't be symmetric, there is a prefix byte that does a thing.

[indutny]

I'm going to work on it.

[indutny]

See #625

[indutny]

Shall be fixed in 87e62bd