JavaScript: How to generate Rfc2898DeriveBytes like C#?

asp.net-identity, c#, cryptojs, javascript, rfc2898

Solution

Ok, I think this problem ended up being quite a bit simpler than I was making it (aren't they always). After performing a RTFM operation on the pbkdf2 spec, I ran some side-by-side tests with Node crypto and .NET crypto, and have made pretty good progress on a solution.

The following JavaScript code correctly parses the stored salt and subkey, then verifies the given password by hashing it with the stored salt. There are doubtless better / cleaner / more secure tweaks, so comments welcome.

// NodeJS implementation of crypto, I'm sure google's 
// cryptoJS would work equally well.
var crypto = require('crypto');

// The value stored in [dbo].[AspNetUsers].[PasswordHash]
var hashedPwd = "ADOEtXqGCnWCuuc5UOAVIvMVJWjANOA/LoVy0E4XCyUHIfJ7dfSY0Id+uJ20DTtG+A==";
var hashedPasswordBytes = new Buffer(hashedPwd, 'base64');

var hexChar = ["0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "A", "B", "C", "D", "E", "F"];

var saltString = "";
var storedSubKeyString = "";

// build strings of octets for the salt and the stored key
for (var i = 1; i < hashedPasswordBytes.length; i++) {
    if (i > 0 && i <= 16) {
        saltString += hexChar[(hashedPasswordBytes[i] >> 4) & 0x0f] + hexChar[hashedPasswordBytes[i] & 0x0f]
    }
    if (i > 0 && i > 16) {
        storedSubKeyString += hexChar[(hashedPasswordBytes[i] >> 4) & 0x0f] + hexChar[hashedPasswordBytes[i] & 0x0f];
    }
}

// password provided by the user
var password = 'welcome1';

// TODO remove debug - logging passwords in prod is considered 
// tasteless for some odd reason
console.log('cleartext: ' + password);
console.log('saltString: ' + saltString);
console.log('storedSubKeyString: ' + storedSubKeyString);

// This is where the magic happens. 
// If you are doing your own hashing, you can (and maybe should)
// perform more iterations of applying the salt and perhaps
// use a stronger hash than sha1, but if you want it to work
// with the [as of 2015] Microsoft Identity framework, keep
// these settings.
var nodeCrypto = crypto.pbkdf2Sync(new Buffer(password), new Buffer(saltString, 'hex'), 1000, 256, 'sha1');

// get a hex string of the derived bytes
var derivedKeyOctets = nodeCrypto.toString('hex').toUpperCase();

console.log("hex of derived key octets: " + derivedKeyOctets);

// The first 64 bytes of the derived key should
// match the stored sub key
if (derivedKeyOctets.indexOf(storedSubKeyString) === 0) {
    console.info("passwords match!");
} else {
    console.warn("passwords DO NOT match!");
}

Problem

EDIT: Per discussion in the comments, let me clarify that this will be happening server side, behind SSL. I do not intend to expose the hashed password or the hashing scheme to the client. Assume we have an existing asp.net identity database with the default tables (aspnet_Users, aspnet_Roles, etc.). Based on my understanding, the password hashing algorithm uses sha256 and stores the salt + (hashed password) as a base64 encoded string. EDIT: This assumption is incorrect, see answer below. I would like to replicate the function of the Microsoft.AspNet.Identity.Crypto class' VerifyHashedPassword function with a JavaScript version. Let's say that a password is welcome1 and its asp.net hashed password is ADOEtXqGCnWCuuc5UOAVIvMVJWjANOA/LoVy0E4XCyUHIfJ7dfSY0Id+uJ20DTtG+A== So far I have been able to reproduce the parts of the method that get the salt and the stored sub key. Where the C# implementation does more or less this: ``` var salt = new byte[SaltSize]; Buffer.BlockCopy(hashedPasswordBytes, 1, salt, 0, SaltSize); var storedSubkey = new byte[PBKDF2SubkeyLength]; Buffer.BlockCopy(hashedPasswordBytes, 1 + SaltSize, storedSubkey, 0, PBKDF2SubkeyLength); ``` I have the following in JavaScript (not elegant by any stretch): ``` var hashedPwd = "ADOEtXqGCnWCuuc5UOAVIvMVJWjANOA/LoVy0E4XCyUHIfJ7dfSY0Id+uJ20DTtG+A=="; var hashedPasswordBytes = new Buffer(hashedPwd, 'base64'); var saltbytes = []; var storedSubKeyBytes = []; for(var i=1;i<hashedPasswordBytes.length;i++) { if(i > 0 && i <= 16) { saltbytes.push(hashedPasswordBytes[i]); } if(i > 0 && i >16) { storedSubKeyBytes.push(hashedPasswordBytes[i]); } } ``` Again, it ain't pretty, but after running this snippet the saltbytes and storedSubKeyBytes match byte for byte what I see in the C# debugger for salt and storedSubkey. Finally, in C#, an instance of Rfc2898DeriveBytes is used to generate a new subkey based on the salt and the password provided, like so: ``` byte[] generatedSubkey; using (var deriveBytes = new Rfc2898DeriveBytes(password, salt, PBKDF2IterCount)) { generatedSubkey = deriveBytes.GetBytes(PBKDF2SubkeyLength); } ``` This is where I'm stuck. I have tried others' solutions such as this one, I have used Google's and Node's CryptoJS and crypto libraries respectively, and my output never generates anything resembling the C# version. (Example: ``` var output = crypto.pbkdf2Sync(new Buffer('welcome1', 'utf16le'), new Buffer(parsedSaltString), 1000, 32, 'sha256'); console.log(output.toString('base64')) ``` generates "LSJvaDM9u7pXRfIS7QDFnmBPvsaN2z7FMXURGHIuqdY=") Many of the pointers I've found online indicate problems involving encoding mismatches (NodeJS / UTF-8 vs. .NET / UTF-16LE), so I've tried encoding using the default .NET encoding format but to no avail. Or I could be completely wrong about what I assume these libraries are doing. But any pointers in the right direction would be much appreciated.

Original source

Related problems