Question

This article explains why string Encoding shouldn’t be used to round-trip the cipher text. However, if I try using recommended Convert.ToBase64String, I get an exception if the original string is not packaged with 4-byte blocks. In the following example, the original string doesn’t work, but just “Zoidberg” does, since it’s 8 bytes long (that are packaged into a 6-byte array when Convert.ToBase64String is used).

As promised in the article, if I use any string Encoding, I get “Bad Data” error when decrypting the value back. So how should the original string be fed into the crypto APIs as a byte array?

string text = "Zoidberg is important!";
RSACryptoServiceProvider provider = new RSACryptoServiceProvider();
// This throws FormatException: "The input is not a valid Base-64 string as it 
// contains a non-base 64 character,  more than two padding characters, 
// or a non-white space character among the padding characters"
byte[] cipherText = provider.Encrypt(Convert.FromBase64String(text), false);
string plainText = Convert.ToBase64String(provider.Decrypt(cipherText, false));

Answer 1

I think you were mislead by that article.
It does not make sense to convert from string -> byte[] --> encrypted byte[] --> string, which is what that article is showing you to not do.

Presumably, you want to go string -> byte[] -> encrypted byte[] -> (network, whatever) -> encrypted byte[] -> byte[] -> string. It’s fine to use Encoding to convert the string to bytes in this manner.

I don’t know why the article you linked would warn you against taking the encrypted bytes and converting that directly back to a string. Seems silly.

You say you’re trying to round-trip the ciphertext, so I would do so with byte arrays containing the encrypted bytes of your string and forget about base-64 encoding.