Get unlimited public & private packages + package-based permissions with npm Pro.Get started »

secrets.js-grempe

2.0.0 • Public • Published

secrets.js

About

secrets.js is an implementation of Shamir's threshold secret sharing scheme in JavaScript, for Node.js and browsers with both Global variable and AMD module loading support.

It can be used to split any "secret" (i.e. a password, text file, Bitcoin private key, anything) into n number of "shares" (each the same size in bits as the original secret), requiring that exactly any number t ("threshold") of them be present to reconstruct the original secret.

This is a fork of the original excellent code created by amper5and on Github. The original secrets.js can be found there.

Security Audit

This code is provided without warranty, as-is and you use it at your own risk.

This library was included in the scope of a security audit performed by Cure53 in July 2019 at the request of the Slant PrivEOS project and no issues were found in this implementation. Our thanks to that team for including this code in its audit scope and for sharing the results with us. A copy of the audit can be found online and in the audit folder of this repository.

The audit was performed on released code on the master branch no later than commit b6e13cb43f2065a9b622a35002a68222f2bfd437 (Wed Aug 3 21:30:11 2016 -0700).

Quoting from the Cure53 audit report:

Conclusion
 
...
 
It can be clarified that PrivEOS is based on the cryptographic algorithm called Shamir’s
Secret Sharing and has been implemented with an EOS smart contract.
 
...
 
Next up, the aforementioned implementation of Shamir's Secret Sharing was audited
and proven to correctly adhere to its specification. Similarly good verdict was reached
about distribution of Shamir’s Secret Sharing.
 

Examples

Divide a 512-bit key, expressed in hexadecimal form, into 10 shares, requiring that any 5 of them are necessary to reconstruct the original key:

// generate a 512-bit key
var key = secrets.random(512) // => key is a hex string
 
// split into 10 shares with a threshold of 5
var shares = secrets.share(key, 10, 5)
// => shares = ['801xxx...xxx','802xxx...xxx','803xxx...xxx','804xxx...xxx','805xxx...xxx']
 
// combine 4 shares
var comb = secrets.combine(shares.slice(0, 4))
console.log(comb === key) // => false
 
// combine 5 shares
comb = secrets.combine(shares.slice(4, 9))
console.log(comb === key) // => true
 
// combine ALL shares
comb = secrets.combine(shares)
console.log(comb === key) // => true
 
// create another share with id 8
var newShare = secrets.newShare(8, shares) // => newShare = '808xxx...xxx'
 
// reconstruct using 4 original shares and the new share:
comb = secrets.combine(shares.slice(1, 5).concat(newShare))
console.log(comb === key) // => true

Divide a password containing a mix of numbers, letters, and other characters, requiring that any 3 shares must be present to reconstruct the original password:

var pw = "<<PassWord123>>"
 
// convert the text into a hex string
var pwHex = secrets.str2hex(pw) // => hex string
 
// split into 5 shares, with a threshold of 3
var shares = secrets.share(pwHex, 5, 3)
 
// combine 2 shares:
var comb = secrets.combine(shares.slice(1, 3))
 
//convert back to UTF string:
comb = secrets.hex2str(comb)
console.log(comb === pw) // => false
 
// combine 3 shares:
comb = secrets.combine([shares[1], shares[3], shares[4]])
 
//convert back to UTF string:
comb = secrets.hex2str(comb)
console.log(comb === pw) // => true

There are some additional examples of simple usage in the browser, Node.js, and AMD loading (require.js) in the examples folder.

Installation and usage

This fork of secrets.js is available from www.npmjs.com. Install using

npm install secrets.js-grempe

The source code for this package is available on Github.

To use it in a Node.js application (Requires OpenSSL support compiled into Node):

var secrets = require("secrets.js")

To use it in the browser with the global 'secrets' defined, include secrets.js or secrets.min.js in your HTML.

<script src="secrets.min.js"></script>

You can also use it in the browser with an AMD module loading tool like require.js. See the AMD loading example in the examples dir.

API

  • secrets.share()
  • secrets.combine()
  • secrets.newShare()
  • secrets.init()
  • secrets.getConfig()
  • secrets.extractShareComponents()
  • secrets.setRNG()
  • secrets.random()
  • secrets.str2hex()
  • secrets.hex2str()

secrets.share( secret, numShares, threshold, [padLength] )

Divide a secret expressed in hexadecimal form into numShares number of shares, requiring that threshold number of shares be present for reconstructing the secret;

  • secret: String, required: A hexadecimal string.
  • numShares: Number, required: The number of shares to compute. This must be an integer between 2 and 2^bits-1 (see secrets.init() below for explanation of bits).
  • threshold: Number, required: The number of shares required to reconstruct the secret. This must be an integer between 2 and 2^bits-1 (see secrets.init() below for explanation of bits).
  • padLength: Number, optional, default 128: How much to zero-pad the binary representation of secret. This ensures a minimum length for each share. See "Note on security" below.

The output of secrets.share() is an Array of length numShares. Each item in the array is a String. See Share format below for information on the format.

secrets.combine( shares )

Reconstructs a secret from shares.

  • shares: Array, required: An Array of shares. The form is equivalent to the output from secrets.share().

The output of secrets.combine() is a String representing the reconstructed secret. Note that this function will ALWAYS produce an output String. However, if the number of shares that are provided is not the threshold number of shares, the output will not be the original secret. In order to guarantee that the original secret is reconstructed, the correct threshold number of shares must be provided.

Note that using more than the threshold number of shares will also result in an accurate reconstruction of the secret. However, using more shares adds to computation time.

secrets.newShare( id, shares )

Create a new share from the input shares.

  • id: Number or String, required: A Number representing the share id. The id is an integer between 1 and 2^bits-1. It can be entered as a Number or a number String expressed in hexadecimal form.
  • shares: Array, required: The array of shares (in the same format as outputted from secrets.share()) that can be used to reconstruct the original secret.

The output of secrets.newShare() is a String. This is the same format for the share that secrets.share() outputs. Note that this function ALWAYS produces an output String. However, as for secrets.combine(), if the number of shares that are entered is not the threshold number of shares, the output share will not be a valid share (i.e. will not be useful in reconstructing the original secret). In order to guarantee that the share is valid, the correct threshold number of shares must be provided.

secrets.init( [bits, rngType] )

Set the number of bits to use for finite field arithmetic.

  • bits: Number, optional, default 8: An integer between 3 and 20. The number of bits to use for the Galois field.
  • rngType: String, optional: A string that has one of the values ["nodeCryptoRandomBytes", "browserCryptoGetRandomValues"]. Setting this will try to override the RNG that would be selected normally based on feature detection. Warning: You can specify a RNG that won't actually work in your environment.

Internally, secrets.js uses finite field arithmetic in binary Galois Fields of size 2^bits. Multiplication is implemented by the means of log and exponential tables. Before any arithmetic is performed, the log and exp tables are pre-computed. Each table contains 2^bits entries.

bits is the limiting factor on numShares and threshold. The maximum number of shares possible for a particular bits is (2^bits)-1 (the zeroth share cannot be used as it is the secret by definition.). By default, secrets.js uses 8 bits, for a total 2^8-1 = 255 possible number of shares. To compute more shares, a larger field must be used. To compute the number of bits you will need for your numShares or threshold, compute the log-base2 of (numShares+1) and round up, i.e. in JavaScript: Math.ceil(Math.log(numShares+1)/Math.LN2). You can examine the current calculated maxShares value by calling secrets.getConfig() and increase the bits accordingly for the number of shares you need to generate.

Note:

  • You can call secrets.init() anytime to reset all internal state and re-initialize.
  • secrets.init() does NOT need to be called if you plan on using the default of 8 bits. It is automatically called on loading the library.
  • The size of the exp and log tables depends on bits (each has 2^bits entries). Therefore, using a large number of bits will cause a slightly longer delay to compute the tables.
  • The theoretical maximum number of bits is 31, as JavaScript performs bitwise operations on 31-bit numbers. A limit of 20 bits has been hard-coded into secrets.js, which can produce 1,048,575 shares. secrets.js has not been tested with this many shares, and it is not advisable to go this high, as it may be too slow to be of any practical use.
  • The Galois Field may be re-initialized to a new setting when secrets.newShare() or secrets.combine() are called with shares that are from a different Galois Field than the currently initialized one. For this reason, use secrets.getConfig() to check what the current bits setting is.

secrets.getConfig()

Returns an Object with the current configuration. Has the following properties:

  • bits: [Number] The number of bits used for the current initialized finite field
  • radix: [Number] The current radix (Default: 16)
  • maxShares: [Number] The max shares that can be created with the current bits. Computed as Math.pow(2, config.bits) - 1
  • hasCSPRNG: [Boolean] Indicates whether or not a Cryptographically Secure Pseudo Random Number Generator has been found and initialized.
    • typeCSPRNG: [String] Indicates which random number generator function has been selected based on either environment feature detection (the default) or by manually specifying the RNG type using secrets.init() or secrets.setRNG(). The current possible types that can be displayed here are ["nodeCryptoRandomBytes", "browserCryptoGetRandomValues"].

secrets.extractShareComponents( share )

Returns an Object with the extracted parts of a public share string passed as an argument. Has the following properties:

  • bits: [Number] The number of bits configured when the share was created.
  • id: [Number] The ID number associated with the share when created.
  • data: [String] A hex string of the actual share data.

secrets.setRNG( function(bits){} | rngType )

Set the pseudo-random number generator used to compute shares.

secrets.js uses a PRNG in the secrets.share() and secrets.random() functions. By default, it tries to use a cryptographically strong PRNG. In Node.js this is crypto.randomBytes(). In browsers that support it, it is crypto.getRandomValues() (using typed arrays, which must be supported too).

To supply your own PRNG, use secrets.setRNG(). It expects a Function of the form function(bits){}. It should compute a random integer between 1 and 2^bits-1. The output must be a String of length bits containing random 1's and 0's (cannot be ALL 0's). When secrets.setRNG() is called, it tries to check the PRNG to make sure it complies with some of these demands, but obviously it's not possible to run through all possible outputs. So make sure that it works correctly.

  • rngType: String, optional: A string that has one of the values ["nodeCryptoRandomBytes", "browserCryptoGetRandomValues"]. Setting this will try to override the RNG that would be selected normally based on feature detection. Warning: You can specify a RNG that won't actually work in your environment.

secrets.random( bits )

Generate a random bits length string, and output it in hexadecimal format. bits must be an integer greater than 1.

secrets.str2hex( str, [bytesPerChar] )

Convert a UTF string str into a hexadecimal string, using bytesPerChar bytes (octets) for each character.

  • str: String, required: A UTF string.
  • bytesPerChar: Number, optional, default 2. The maximum bytesPerChar is 6 to ensure that each character is represented by a number that is below JavaScript's 2^53 maximum for integers.

secrets.hex2str( str, [bytesPerChar] )

Convert a hexadecimal string into a UTF string. Each character of the output string is represented by bytesPerChar bytes in the String str. See note on bytesPerChar under secrets.str2hex() above.

Share Format

Each share is a string in the format <bits><id><value>. Each part of the string is described below:

  • bits: The first character, expressed in Base36 format, is the number of bits used for the Galois Field. This number must be between 3 and 20, expressed by the characters [3-9, a-k] in Base36.
  • id: The id of the share. This is a number between 1 and 2^bits-1, expressed in hexadecimal form. The number of characters used to represent the id is the character-length of the representation of the maximum id (2^bits-1) in hexadecimal: (Math.pow(2,bits)-1).toString(16).length.
  • data: The value of the share, expressed in hexadecimal form. The length of this string depends on the length of the secret.

You can extract these attributes from a share in your possession with the secrets.extractShareComponents(share) function which will return an Object with these attributes. You may use these values, for example, to call secrets.init() with the proper bits setting for shares you want to combine.

Note on Security

Shamir's secret sharing scheme is "information-theoretically secure" and "perfectly secure" in that less than the requisite number of shares provide no information about the secret (i.e. knowing less than the requisite number of shares is the same as knowing none of the shares). However, because the size of each share is the same as the size of the secret (when using binary Galois fields, as secrets.js does), in practice it does leak some information, namely the size of the secret. Therefore, if you will be using secrets.js to share short password strings (which can be brute-forced much more easily than longer ones), it would be wise to zero-pad them so that the shares do not leak information about the size of the secret. With this in mind, secrets.js will zero-pad in multiples of 128 bits by default which slightly increases the share size for small secrets in the name of added security. You can increase or decrease this padding manually by passing the padLength argument to secrets.share().

When secrets.share() is called with a padLength, the secret is zero-padded so that it's length is a multiple of the padLength. The second example above can be modified to use 1024-bit zero-padding, producing longer shares:

var pw = "<<PassWord123>>"
 
// convert the text into a hex string
var pwHex = secrets.str2hex(pw) // => 240-bit password
 
// split into 5 shares, with a threshold of 3, WITH zero-padding
var shares = secrets.share(pwHex, 5, 3, 1024) // => 1024-bit padded shares
 
// combine 3 shares
var comb = secrets.combine([shares[1], shares[3], shares[4]])
 
// convert back to UTF string
comb = secrets.hex2str(comb)
 
console.log(comb === pw) // => true

There is a workbook containing some additional padding examples.

License

secrets.js is released under the MIT License. See the LICENSE file.

Development and Testing

cd secrets.js/
npm install

Continuous Development

Watch all JavaScript files and run the testing and minification tasks on every save to a file.

npm run watch

Build & Minify

The minified version of the secrets.js can be found in secrets.min.js. This file was generated using the UglifyJS2 tool.

npm run build

Node.js Testing with Jasmine

You can also run the Jasmine test suite within a Node.js instance.

npm run test

Browser Testing with Jasmine

There is a Jasmine test suite that exercises the entire secrets module that can be run against the source code or the minified version.

npm run test-browser
npm run test-browser-min

Changelog

  • 2.0.0

    • [BREAKING] Removed SJCL random number generator. Modern browsers have all the support they need for crypto.getRandomValues().
    • Modernize build and test process
    • Lint and Prettier JS and Markdown files
    • Implement npm run ... scripts for common tests and build tasks
    • Remove global dev dependencies
    • Remove references to, and use of, Bower installation
  • 1.2.0

    • Added secrets.seedRNG() function to allow seeding the SJCL RNG instantly via Browser or Node.js RNG's or with entropy from an external server.
  • 1.1.0

    • Added grunt watch task to auto-run tests and minification on every JavaScript file save.
    • Minified file now contains name, version and author comments automatically.
    • Configured basic grunt tasks for minification, Node.js testing with Jasmine, jshint, eslint. Removed Karma test runner and manual minification and testing steps. Just run grunt.
    • [Bugfix] calling secrets.init() now actually resets all internal state back to the default settings. Previously init() only reset some internal values. init() now calls a new private function reset() to accomplish this.
    • [Enhancement] If the Stanford Javascript Crypto Libarary (SJCL) is loaded in the browser it can be used as a fallback, or explicitly selected, CSPRNG for those browsers that don't support crypto.getRandomValues(). It uses the Fortuna RNG and collects additional entropy from mouse movements continually. The downside is that it requires mouse movements initially before secrets.random() can be called. secrets.random() will throw an Error if called when SJCL is not fully seeded. Currently set to use the maximum SJCL 'paranoia' level of 10. An enhancement to this might be to call out to retrieve one or more external sources of entropy (and mixing them together) to pre-seed the RNG when the library is loaded.
    • [Enhancement] You can now pass a string to init() or setRNG() which forces loading of a specific RNG (whether it will work or not in your current env!) * Re-factored how getRNG() works internally. Now it returns small focused functions, not a giant function with detection conditionals. If SJCL is loaded the RNG tests are skipped since they would always initially fail due to the entropy pool being initally empty. This should be OK for this 'trusted' RNG.
  • 1.0.0

    • Packaging cleanup and ready for 1.0.0 release on Bower and NPM.
    • [Enhancement] Now supports the Javascript Universal Module Definition UMDJS for loading this module in the Browser with a secrets global, using an AMD Module loader like require.js, or in Node.js apps.
    • Refactor getRNG() to no longer have embedded require now that crypto is included on module load with the UMDJS change.
    • Updated README.md with info about this fork of secrets.js. * Added some simple examples of usage to the examples folder.
  • 0.2.0

    • [Enhancement] Extend the output of getConfig() to include the radix and maxShares properties.
    • [Security] Zero-pad all secrets in multiples of 128 bits (instead of 0) by default.
    • [Performance] Massive (100x) speed optimization to padLeft() private function (the second most frequently called block of code internally).
    • [Testing] Added a full jasmine test suite and Karma test runner. Karma runs will also generate code coverage HTML reports. Code coverage is currently >90%.
    • [Testing] Expose all private functions as Underscore (_) prefixed functions to allow direct unit testing.
    • [Security] Removed Math.random fallback random number generator. Should always fail safe, even if it means not working. secrets.getConfig().unsafePRNG will always result in undefined now as it is no longer ever set.
    • Refactored away need to know anything about global var.
    • [Testing] jslint.com, jshint.com, and eslint CLI warnings for code and style now clean. - Beautify code.
  • 0.1.8: bugfix release

  • 0.1.7: added config.unsafePRNG reset when supplying a new PRNG

  • 0.1.6:

    • Removed JSBN dependency, support for arbitrary radices, and the convertBase() function, with attendant 50% file size reduction.
    • Fixed bug where leading zeros were dropped. * Renamed string conversion functions.
  • 0.1.5: getConfig() returns information about PRNG

  • 0.1.4: new share format

Install

npm i secrets.js-grempe

DownloadsWeekly Downloads

4,467

Version

2.0.0

License

MIT

Unpacked Size

1.46 MB

Total Files

29

Last publish

Collaborators

  • avatar