ConvertToUTF8 | Sublime Text 2 & 3 plugin | Code Editor library

Your code makes no sense: converting a String to UTF8 and getting back a String makes no sense: a String contains characters. Not bytes.

Encoding a key to base64 doesn't make much sense either. Encoding the plain text to base64 is useless, too.

You need base64 encoding when you have random, binary bytes, and you want to transform them to printable english characters. Only then.

So the process should be:

1. transform your key to bytes (using String.getBytes(UTF_8)). Unless the String key is in fact a base64-encoded byte array, in which case you need to base64 decode it;
2. transform the plain text to bytes (using String.getBytes(UTF_8));
3. encrypt the bytes you got from step 2, using the key obtained from step 1. You obtain completely "random" bytes. These bytes don't represent characters encoded in your platform default charset. So transforming them to a String using new String(bytes) doesn't make any sense, and is a lossy transformation.
4. Just return the result as a byte array, or if you really want printable characters, base64-encode the bytes, and return the string you obtain from this base64 encoding.

To decrypt, use the reverse process:

1. Use the same thing as in step 1 above to get the key
2. take the bytes obtained from step4 above (if you chose to return a byte array), or base64-decode the string, to obtain the original random binary bytes
3. decrypt the bytes obtained from step 2, using the key obtained from step 1. You get a byte array representing the UTF8-encoded characters of the original plain text.
4. Use new String(decryptedBytes, UTF_8) to transform this byte array to a String.

You won't get a always working version without converting your project into a Unicode aware application.

In other words, to support all chars that may part inside a BSTR you need a Unicode CString (CStringW)

You may stay with an MBCS version but in this case you still have to handle Unicode. Using CStringW may be an option here.

Converting to UTF-8 is done with WideCHarToMultiByte

Assigning an encryption with a String object is a common mistake. Strings are supposed to be an abstraction over such implementation details as encoding. As long as you're only working with Strings, all you need to know is that they're Unicode. The specific encoding comes into play only when you convert String to bytes, which in your case will probably mean in some HTTP client. I don't know what do you use, so I've attached an example for the one from Apache.