I’m trying to support PBE for AES, Serpent, and TwoFish. Currently I am able to generate an AES PBEKey in Java using BC like this:
SecretKeyFactory factory = SecretKeyFactory.getInstance("PBEWITHSHA256AND256BITAES-CBC-BC", provider);
PBEKeySpec pbeKeySpec = new PBEKeySpec("Password12".toCharArray());
SecretKey key = factory.generateSecret(pbeKeySpec);
but I can’t figure out how to generate a PBEKey for Serpent, so I’m assuming its not possible out of the box. How would I go about implementing this? Is there a hook somewhere that I can just register my own SecretKeyFactory to handle Serpent keys?
Coincidentally, I have noticed that using an AES PBEKey (as generated above) for encrypting/decrypting with Serpent/TwoFish “works”, but I have no idea what the repercussions are. Could I just get away with using the AES PBEKey?
After discussions with PaŭloEbermann (above), I put together the following solution. It generates a PBE key for AES256 and then simply copies the required number of bytes from the generated key into a new SecretKeySpec(), which allows me to specify the desired algorithm and key length. Currently I am salting the password AND creating a random IV on each call to encrypt. My assumption is that the IV is unnecessary since a random salt is applied to each encrypted message, but I wasn’t 100% sure so I added the IV anyway. I’m hoping someone can confirm or deny this assumption, since if the IV isnt needed, then its bloating the size of the output from encrypt() for no valid reason. Ideally I would be able to generate a PBEKey of variable length with no algorithm ties (as per PKCS5), but it appears I am bound to the key sizes defined in the available PBE schemes provided by the selected Provider. This implementation is therefore bound to using BouncyCastle, since I was unable to find a PBE scheme that provided at least 256bit keys from the standard JCE provider.