• Fixed TypeScript typings (#157)
• Rebuilt using newer version of Uglify-js.

🗝 🎉 🔐

Changes since v1.0.0-rc.1

No code changes

Changes since v0.14.5:

• IMPORTANT! In previous versions, nacl.secretbox.open, nacl.box.open, and nacl.box.after returned false when opening failed (for example, when using incorrect key, nonce, or when input was maliciously or accidentally modified after encryption). This version instead returns null.

  The usual way to check for this condition:

  if (!result) { ... }

  is correct and will continue to work.

  However, direct comparison with false:

  if (result == false) { ... }

  it will no longer work and will not detect failure. Please check your code for this condition.

  (nacl.sign.open always returned null, so it is not affected.)

• Arguments type check now uses instanceof Uint8Array instead of Object.prototype.toString.

• Removed deprecation checks for nacl.util (moved to a
  separate package in v0.14.0).

• Removed deprecation checks for the old signature API (changed in v0.10.0).

• Improved benchmarking.

• IMPORTANT! In previous versions, nacl.secretbox.open, nacl.box.open, and nacl.box.after returned false when opening failed (for example, when using incorrect key, nonce, or when input was maliciously or accidentally modified after encryption). This version instead returns null.

  The usual way to check for this condition:

  if (!result) { ... }

  is correct and will continue to work.

  However, direct comparison with false:

  if (result == false) { ... }

  it will no longer work and will not detect failure. Please check your code for this condition.

  (nacl.sign.open always returned null, so it is not affected.)

• Arguments type check now uses instanceof Uint8Array instead of Object.prototype.toString.

• Removed deprecation checks for nacl.util (moved to a separate package in v0.14.0).

• Removed deprecation checks for the old signature API (changed in v0.10.0).

• Improved benchmarking.

• Fixed incomplete return types in TypeScript typings.
• Replaced COPYING.txt with LICENSE file, which now has public domain dedication
  text from The Unlicense. License fields in package.json and bower.json have
  been set to "Unlicense". The project was and will be in the public domain --
  this change just makes it easier for automated tools to know about this fact by
  using the widely recognized and SPDX-compatible template for public domain
  dedication.

• Added TypeScript type definitions (contributed by @AndSDev).
• Improved benchmarking code.

The bug in the fast version of Poly1305 has been fixed and this version is back into nacl-fast.js. Thanks to @floodyberry for promptly responding and fixing it:

"The issue was not properly detecting if st->h was >= 2^130 - 5, coupled with [testing mistake] not catching the failure. The chance of the bug affecting anything in the real world is essentially zero luckily, but it's good to have it fixed."

https://github.com/floodyberry/poly1305-donna/issues/2#issuecomment-202698577

Commit in TweetNaCl.js: https://github.com/dchest/tweetnacl-js/commit/6dcbcaf5f5cbfd313f2dcfe763db35c828c8ff5b Commit in Poly1305-donna: https://github.com/floodyberry/poly1305-donna/commit/0911057b9607f37e642337739a2834aec84fbb53

This update switches Poly1305 fast version back to original (slow) version.

There was a bug in fast version of Poly1305 which sometimes produced incorrect results. Authenticity of results of the following functions may be affected: secretbox, secretbox.open, box, box.open.

Fast version is default if you used npm package (e.g. require("tweetnacl")).

TweetNaCl.js comes in two favors: nacl.js, which is almost the exact port of TweetNaCl, and nacl-fast.js, which includes faster versions ported from other implementations. The fast version of Poly1305 message authenticator comes from 16-bit version of floodyberry/poly1305-donna. The bug is present in this version and was ported to JavaScript.

Until we figure how to fix the bug, Poly1305 in nacl-fast.js was switched to original nacl.js.

No code changes, just tweaked packaging and added COPYING.txt.

• Breaking change! All functions from nacl.util have been removed. These functions are no longer available:

  nacl.util.decodeUTF8
  nacl.util.encodeUTF8
  nacl.util.decodeBase64
  nacl.util.encodeBase64
  

  If want to continue using them, you can include https://github.com/dchest/tweetnacl-util-js package:

  <script src="nacl.min.js"></script>
  <script src="nacl-util.min.js"></script>
  

  or

  var nacl = require('tweetnacl');
  nacl.util = require('tweetnacl-util');
  

  However it is recommended to use better packages that have wider compatibility and better performance. Functions from nacl.util were never intended to be robust solution for string conversion and were included for convenience: cryptography library is not the right place for them.

  Currently calling these functions will throw error pointing to tweetnacl-util-js (in the next version this error message will be removed).

• Improved detection of available random number generators, making it possible to use nacl.randomBytes and related functions in Web Workers without changes.

• Changes to testing (see README).

No code changes.

• Reverted license field in package.json to "Public domain".
• Fixed typo in README.

• Fixed undefined variable bug in fast version of Poly1305. No worries, this bug was never triggered.
• Specified CC0 public domain dedication.
• Updated development dependencies.

• Exclude crypto and buffer modules from browserify builds.

• Made nacl-fast the default version in NPM package. Now require("tweetnacl") will use fast version; to get the original version, use require("tweetnacl/nacl.js").
• Cleanup temporary array after generating random bytes.

• Improved performance of curve operations, making nacl.scalarMult, nacl.box, nacl.sign and related functions up to 3x faster in nacl-fast version.

• Significantly improved performance of Salsa20 (~1.5x faster) and Poly1305 (~3.5x faster) in nacl-fast version.

• Instead of using the given secret key directly, TweetNaCl.js now copies it to a new array in nacl.box.keyPair.fromSecretKey and nacl.sign.keyPair.fromSecretKey.

• Added new constant: nacl.sign.seedLength.

• Even faster hash for both short and long inputs (in nacl-fast).

• Implement nacl.sign.keyPair.fromSeed to enable creation of sign key pairs deterministically from a 32-byte seed. (It behaves like libsodium's crypto_sign_seed_keypair: the seed becomes a secret part of the secret key.)
• Fast version now has an improved hash implementation that is 2x-5x faster.
• Fixed benchmarks, which may have produced incorrect measurements.

• Exported undocumented nacl.lowlevel.crypto_core_hsalsa20.

• Signature API breaking change! nacl.sign and nacl.sign.open now deal with signed messages, and new nacl.sign.detached and nacl.sign.detached.verify are available.

  Previously, nacl.sign returned a signature, and nacl.sign.open accepted a message and "detached" signature. This was unlike NaCl's API, which dealt with signed messages (concatenation of signature and message).

  The new API is:

  nacl.sign(message, secretKey) -> signedMessage
  nacl.sign.open(signedMessage, publicKey) -> message | null
  

  Since detached signatures are common, two new API functions were introduced:

  nacl.sign.detached(message, secretKey) -> signature
  nacl.sign.detached.verify(message, signature, publicKey) -> true | false
  

  (Note that it's verify, not open, and it returns a boolean value, unlike open, which returns an "unsigned" message.)

• NPM package now comes without test directory to keep it small.

• Improved documentation.
• Fast version: increased theoretical message size limit from 2^32-1 to 2^52 bytes in Poly1305 (and thus, secretbox and box). However this has no impact in practice since JavaScript arrays or ArrayBuffers are limited to 32-bit indexes, and most implementations won't allocate more than a gigabyte or so. (Obviously, there are no tests for the correctness of implementation.) Also, it's not recommended to use messages that large without splitting them into smaller packets anyway.

Note: The library is stable and API is frozen, however it has not been independently reviewed.