CowCrypt provides popular hashing and encryption algorithms for the MooTools JavaScript framework. My goal is to provide much of the same robust functionality as the crypto-js library, but with the elegance and familiar syntax of MooTools. The code is licensed under the MIT License.
This is an experimental project and the algorithms are not fully optimized for speed. However, many of them perform neck-in-neck with other popular crypto libraries, and some of them are much faster (some of them are slower :p).
Please contact me if you have any questions, issues, or feature requests. Feel free to use the GitHub Issue tracker or email me <jeff@rubbingalcoholic.com>
- WARNING: Encode your strings as ASCII!
- The Hashers: MD5, SHA-1, SHA-256
- HMAC
- Symmetric Encryption: AES, CAST5, Twofish
- RSA Encryption
- PBKDF2
- Contributing
CowCrypt's hashing and encryption functions accept string inputs, but they must be ASCII-encoded (ie. containing no UTF8 / multibyte character code values). The algorithms do not slow down to check whether your strings are actually ASCII. This will lead to broken or corrupted results if you pass multibyte characters. NOTABUG! WONTFIX!
CowCrypt provides convenient conversion methods to encode UTF strings to ASCII, but you are responsible for doing this as needed outside of the encryption and hashing functions. Be careful!
var mystring = 'ʟʘȽ ƮɦɨƧ ȘƬɌȋɴɢ ɪŜ ƜǡƇƘŸǃ ßƐŧŧëƦ ȄƞĈøÐe ĩť ƒĭƦŚȚ.';
var mystring = convert.utf8.encode(mystring);
var hashed = new SHA1().hash(mystring);
CowCrypt includes the following classes for hashing data:
-
MD5: An old hashing algorithm that has fallen out of favor due to its susceptibility to collision attacks. A lot of apps still use it, so here it is.
-
SHA1: Designed to replace MD5. SHA-1 has some theoretical security weaknesses, but it's still the standard and the most widely used hasher.
-
SHA256: Also known as SHA-2. It was designed to address the security problems with SHA-1. Since there's no (known) real world attack on SHA-1, it isn't widely used.
All of these classes extend the Hasher base class, which provides common functionality and makes the usage syntax more or less identical. You can read about this in the API docs. OK, here are the usage examples:
var hash = new MD5().hash('data data data ');
// 760c2be99e98ae3027ae4d4c2816d6ea
var hash = new SHA1().hash('data data data ');
// 65bd90d5e213e8d03e87b5be5eeda3bc81faa772
var hash = new SHA256().hash('data data data ');
// 6b6d03945132109b4e8c035318219e9553f9e772dbf9392094492b2ea8a4e9ad
Normally you pass your data into the hash function and you're done with it. But if you're hashing huge amounts of data, perhaps a Linux ISO, it's a big performance drain to pass all of it around in memory at once. Since the hasher functions really only operate on 64 bytes of data at a time, there's no reason we should have to pass all that data into them in one call.
Using streaming (or progressive hashing) mode, we can stream data into the hasher over multiple calls, potentially reducing our overall memory footprint. See below.
var _sha1 = new SHA1();
for (var i = 0; i < 3; i++)
_sha1.hash('data ', {stream: true});
var hash = _sha1.finalize();
// 65bd90d5e213e8d03e87b5be5eeda3bc81faa772
By default, each of the hashers returns a hexadecimal string. You can change this to an ASCII-encoded binary string, or an array of 32-bit integer "words" using the return_format option for the hash and finalize methods.
// in normal hashing mode, pass return_format option on hash
var hash = new SHA1().hash('data', {return_format: 'words'});
// in streaming mode, pass return_format option into finalize
var _sha1 = new SHA1();
for (var i = 0; i < 3; i++)
_sha1.hash('data ', {stream: true});
var hash = _sha1.finalize({return_format: 'binary');
HMAC is a method of hashing data that uses a passphrase in the hashing process. This allows you to verify both data integrity and authenticity.
var _hmac = new HMAC({
passphrase: 'password1234lol',
hasher: SHA1
});
_hmac.hash('my data to hash');
This wraps the streaming mode functionality for the Hasher subclass instance used by HMAC. See above for a more complete explanation.
var _hmac = new HMAC({ passphrase: '12345678'});
_hmac.hash('Message Part 1', {stream: true});
_hmac.hash('Message Part 2', {stream: true});
_hmac.hash('Message Part 3', {stream: true});
var output = _hmac.finalize();
So far, CowCrypt has the following classes for symmetric encryption:
-
AES: Also known as Rijndael, AES is effectively the world standard symmetric block cipher. It accepts key sizes of 16, 24, and 32 bytes, so you can choose between speed and moar security.
-
CAST5: This is kind of an obscure algorithm. It's used internally by OpenPGP, so it's good to have around if you were thinking about building a JavaScript OpenPGP application.
-
Twofish: This went head-to-head against Rijndael as a finalist in NIST's AES-selection process. Although it lost the competition, it is a very secure algorithm and the dude who built it (Bruce Schneier) it is a total genius.
All of these classes extend the BlockCipher base class, which provides common functionality and makes the usage syntax more or less identical. You can read about this in the API docs.
// This encrypts by deriving a key from the passphrase 'mypassword1234'
var _aes = new AES({passhprase: 'mypassword1234'});
var data = convert.utf8.encode('i like ŞĩŁĿŶ ǙƝȈʗʘɗε characters');
var encrypted = _aes.encrypt(data);
You can also directly specify a key and initial vector (IV).
var my_key = convert.hex_to_binstring('0123456712345678234567893456789a');
var my_iv = convert.hex_to_binstring('9876543210fedcba');
var _cast5 = new CAST5({key: my_key, iv: my_iv});
var crypted = _cast5.encrypt('ASCII string! no UTF encoding needed! lol');
CowCrypt includes the following block operation mode classes:
- CBC (default)
- CFB
- ECB (weak sauce)
To customize, initialize your cipher with the block_mode option set to a reference of your preferred block operation mode class.
// initialize with ECB mode:
var cipher = new Twofish({passphrase: '1234', block_mode: ECB});
// initialize with CFB mode:
var cipher = new Twofish({passphrase: '1234', block_mode: CFB});
CowCrypt includes the following padding mode classes:
- PKCS7 (default)
- ANSIX923
- ZeroPadding
When your input plaintext doesn't match the ciphers block size, it is padded before encryption according to whatever byte padding mode class is selected. Initialize the cipher with the pad_mode option to override the default. Padding will automatically be removed after decryption, except in the case of Zero Padding, which can't be safely removed without potentially corrupting the data.
// initialize with Zero Padding mode:
var cipher = new CAST5({passphrase: '1234', pad_mode: ZeroPadding});
// initialize with ANSIX923 mode:
var cipher = new Twofish({passphrase: '1234', pad_mode: ANSIX923});
All of the BlockCipher subclasses output ASCII-encoded binary strings for both encryption and decryption. You can easily convert this to other formats using the provided convert library:
var encrypted = new AES({passphrase: '1234'}).encrypt('lol i have herpes');
var hex = convert.binstring_to_hex(encrypted); // convert to hex
var b64 = convert.base64.encode(encrypted); // convert to Base64
OpenSSL uses a special format for its encrypted data which prepends salt information to the ciphertext. CowCrypt offers an OpenSSL compatibility mode for all BlockCipher subclasses. This changes the ciphertext output format and passhprase-based key generation method to allow interoperability with OpenSSL.
Suppose we save the text "this is my encrypted file, hopefully." into a file called in.txt. Then we can use OpenSSL to encrypt with the shell command:
openssl enc -aes-256-cbc -in in.txt -pass pass:"Secret Passphrase" -base64
The output of this command will be some Base64-encoded encrypted data stream using a random salt (different every time). We can decrypt it with CowCrypt.
var openssl_data = convert.base64.decode('U2FsdGVkX18hkhQhtWqrwimjR4BBoZvK\
tdd9MzDuuqU+pyHD+T2aA2/XvB6OY5bvaH7fTvIxGz7Yve01j25+LQ==');
var _aes = new AES({
block_mode: CBC,
pad_mode: PKCS7,
passphrase: 'Secret Passphrase',
openssl_mode: true
});
var plaintext = _aes.decrypt(openssl_data);
// outputs "this is my encrypted file, hopefully."
CowCrypt supports RSA encryption, decryption and key-generation using a FIPS-186-4-compliant probable prime generation algorithm. These features can be run in a separate thread using HTML5 Web Workers for smooth performance. See the Threaded RSA Key Generation and Decryption tutorial for more information on setting this up.
Check out the RSA demo page to see everything in action.
RSA uses huge integers, sometimes upwards of 2048 bits. JavaScript's native support for integers tops out at 53 bits, so we've wrapped up Leemon's awesome BigInt.js library into crypto_math.js to give us those extra bits. Keep in mind:
- 1: crypto_math.js is required for all RSA functionality
- 2: The values used for public and private keys are BigInts, and cannot be treated like normal integers. This should be transparent to you, assuming you use CowCrypt's built in RSA functions.
You can parse a BigInt value from a string of digits as follows:
var big_int = new BigInt().parse("812345834502348952793");
CowCrypt can generate 2048- or 3072-bit RSA keys. The process is CPU-intensive and can take from several seconds to over a minute, depending on the hardware used. To prevent the browser from becoming unresponsive, RSA key generation must be done in a separate thread. This allows the algorithms to work in the background without impacting perceived browser performance. See the Threaded RSA Key Generation and Decryption tutorial for more information.
The non-threaded RSA encryption / decryption methods in the RSA class expect you to pass an RSAKey object. This object holds the BigInt values associated with a public or private key. Private keys have some optional properties that store additional data about how the key was generated. This will be useful for serializing OpenPGP key data packets once support for that is implemented ;)
// The threaded key generation process will generate all these input values
var key = new RSAKey({
n: key_modulus, // required for all keys
e: public_exponent, // required for public keys
d: private_exponent, // required for private keys
p: large_prime_p, // optional for private keys
q: large_prime_q, // optional for private keys
u: mult_inverse_p_mod_q // optional for private keys
});
CowCrypt uses the EME PKCS1 v1.5 encoding method for RSA plaintexts specified in RFC-4880 (the OpenPGP spec). This is can encrypt messages up to 11 bytes less than the byte-length of the modulus, and is useful for symmetric key establishment (a la OpenPGP).
var public_key = {YOUR RSAKey Instance Here};
var plaintext = "Your time is up I will no longer play for you.";
var ciphertext = new RSA().encrypt(plaintext, public_key);
Once you actually have a private key and some ciphertext, decrypting is easy. Note that decrypting is generally a lot slower than encrypting, so this is a good thing to do in a separate Worker thread.
var private_key = {YOUR RSAKey Instance Here};
// generally RSA ciphertext is Base64-encoded. Convert to a binary string!
ciphertext = convert.base64.decode(ciphertext);
var plaintext = new RSA().decrypt(ciphertext, private_key);
If you're using the threaded RSA functions for encryption and decryption, you must use encode your plaintext before encryption, and decode your decrypted plaintext after decryption using the PKCS1 v1.5 class. The threaded functions assume you've already done this, and they won't break, but your ciphertext will be a lot less secure if you don't encode properly. Anyhoo, this is easy:
// BEFORE ENCRYPTING
var encoded_plaintext = new PKCS1_v1_5().encode(plaintext);
// ...
// AFTER DECRYPTING
var decoded_plaintext = new PKCS1_v1_5().decode(encoded_plaintext);
PBKDF2 is a standard for turning a passphrase into a symmetric key using an optional salt. This is used internally by the symmetric ciphers when you initialize with a passphrase instead of explicitly passing a key and initial vector. Note that the PBKDF2 compute method returns an ASCII-encoded binary string, just like how the symmetric ciphers do.
var _kdf = new PBKDF2({
key_size: 32, // note the key size is in bytes
hasher: SHA256, // PBKDF2 uses HMAC internally
iterations: 1000 // moar = bettar (slowar)
});
var key = _kdf.compute('password1234', 'arbitrary salt value');
Please let me know if you have any suggestions for improvements. If you're code savvy, fork the project and make the change yourself! I will do my best to help if something doesn't work or isn't clear. You can find me on Twitter @rubbingalcohol