Java AES/CFB8/NoPadding encryption and Charset

Question

I'm trying to encrypt and decrypt Strings in Java and PHP, so i use AES/CFB8/NoPadding that sould work on both sides.

Now my Cryptor works quit well with Latin Characters, as long as i set the Charset to Latin. But when i set it to UTF-8 (which is what i use in my DB), the encryption is not properly done.

Here is the output:

/*
* Latin (ISO-8859-1) output:
* Original: MiiiMüäöMeeʞ
* Encoded: rQ¶[ÉÐRíD
* Decoded: MiiiMüäöMee?
* 
* UTF-8 output:
* Original: MiiiMüäöMeeʞ
* Encoded: rQ�[�
* Decoded: Mii0SS1])_�ELJI�S�;�W��W?*
*/

Since "ʞ" is not a latin character, I understand it can not be encrypted properly. But why does UTF-8 not work?

public class Cryptor {

    private Cipher cipher;
    private String secretKey = "1234567890qwertz";
    private String iv = "1234567890qwertz";

    private SecretKey keySpec;
    private IvParameterSpec ivSpec;
    private Charset CHARSET = Charset.forName("ISO-8859-1"); // ISO-8859-1 vs. UTF-8

    public Cryptor() throws CryptingException {

        keySpec = new SecretKeySpec(secretKey.getBytes(CHARSET), "AES");
        ivSpec = new IvParameterSpec(iv.getBytes(CHARSET));
        try {
            cipher = Cipher.getInstance("AES/CFB8/NoPadding");
        } catch (NoSuchAlgorithmException e) {
            throw new SecurityException(e);
        } catch (NoSuchPaddingException e) {
            throw new SecurityException(e);
        }
    }

    public String decrypt(String input) throws CryptingException {

        try {
            cipher.init(Cipher.DECRYPT_MODE, keySpec, ivSpec);
            return new String(cipher.doFinal(input.getBytes(CHARSET)), CHARSET).trim();
        } catch (IllegalBlockSizeException e) {
            throw new SecurityException(e);
        } catch (BadPaddingException e) {
            throw new SecurityException(e);
        } catch (InvalidKeyException e) {
            throw new SecurityException(e);
        } catch (InvalidAlgorithmParameterException e) {
            throw new SecurityException(e);
        }
    }

    public String encrypt(String input) throws CryptingException {
        try {
            cipher.init(Cipher.ENCRYPT_MODE, keySpec, ivSpec);
            return new String(cipher.doFinal(input.getBytes(CHARSET)), CHARSET).trim();
        } catch (InvalidKeyException e) {
            throw new SecurityException(e);
        } catch (InvalidAlgorithmParameterException e) {
            throw new SecurityException(e);
        } catch (IllegalBlockSizeException e) {
            throw new SecurityException(e);
        } catch (BadPaddingException e) {
            throw new SecurityException(e);
        }
    }

    public static void main(String Args[]) {

        try {
            Cryptor c = new Cryptor();
            String original = "MiiiMüäöMeeʞ";
            System.out.println("Original: " + original);
            String encrypted = c.encrypt("MiiiMüäöMeeʞ");
            System.out.println("Encoded: " + encrypted);
            System.out.println("Decoded: " + c.decrypt(encrypted));

        } catch (CryptingException e) {
            e.printStackTrace();
        }
    }

    class CryptingException extends RuntimeException {

        private static final long serialVersionUID = 7123322995084333687L;

        public CryptingException() {
            super();
        }

        public CryptingException(String message) {
            super(message);
        }
    }
}

Answer 1

I think turning the encrypted bytes into a String is a bad idea. The bytes are not valid for any encoding, they are random.

You need to encode the resulting byte[] to base64 to get a consistent outcome. See sun.misc.BASE64Encoder/sun.misc.BASE64Decoder.

Here is an example of decoding a base64 String to a byte[], the reverse process is very similar.

You can declare the decoder/encoder and the top of the class:

private final BASE64Decoder base64Decoder = new BASE64Decoder();
private final BASE64Encoder base64Encoder = new BASE64Encoder();

Then in your decypt method you need to call

return new String(cipher.doFinal(base64Decoder.decodeBuffer(input)), CHARSET);

And in your encrypt method

return base64Encoder.encode(cipher.doFinal(input.getBytes(CHARSET)));

Output using UTF-8:

Original: MiiiMüäöMeeʞ
Encoded: clEUtlv2ALXsKYw4ivOfwQ==
Decoded: MiiiMüäöMeeʞ

One side note is that it's not strictly speaking good practice to use packages from sun.* as they are not part of the java spec and hence may change/vanish from version to version.

Here is a little article on moving to the Apache Commons Codec implementation.
This is also a similar class in the Guava API.