The SimpleMembershipProvider, Secure Passwords And The Crypto Helper

Some people have questioned the security of the ASP.NET SimpleMembershipProvider's storage of passwords. The cause for concern seems to stem from the fact that the PasswordSalt field in the standard membership table is unused by the SimpleMembershipProvider, whereas it is used by, for example, SqlMembershipProvider. So what is the PasswordSalt field, and why should the fact that it isn't used raise an eyebrow?

Everyone knows that you should not store user passwords in plain text, right? Doing so is highly irresponsible. It may be that the content you offer is not high value, and that if someone manages to compromise your password store, they won't really profit from using any of the credentials to log in to restricted areas in your site. However, it is a fact that many people reuse passwords across multiple sites, so you may end up providing the malicious user with someone's keys to another kingdom altogether. If you are asking people to provide their personal information to you, then you owe those people a duty of care.

So how should you store passwords? Generally, the accepted way to store a password is not to store it at all. You should instead store a hash. A hash function is an algorithm that takes a block of data, such as a string, and generates a new value, or hash, based on the input. Hashes are fixed size. They are also deterministic. That means that the same value subjected to the same algorithm will always result in the same hash. When a user first registers with your site, you take their proffered password, hash it and store the hash in a database. Whenever they log in after that, you take their submitted password, hash it and compare that hash to the one stored in the database.

Hashing is one-way. It should be infeasible to compute the original value from a given hash. People have a habit of using real words and names as passwords. The chances of duplicate passwords grows when real words are used, and since hashing is deterministic, the hash will be identical for those users that share the same password. The presence of duplicate hashes in a password table is a good indicator that real words or names were used. Hackers can use a number of techniques to attempt to crack hashed passwords, including Brute Force, Dictionary Attacks and Rainbow Tables. All of these approaches are ineffective against hashes that include "salts".

In cryptographic terms, a salt is a random value that is concatenated with the password before hashing it. If you salt your hashes, the salt itself needs to be included with the password prior to hashing it for comparison with the stored value. Now you begin to see why people are concerned about the SimpleMembershipProvider - since the PasswordSalt field in the database is not used, it looks like password hashes are not salted and are therefore not as secure as they should be.

The good news is that appearances can be deceptive, and that the SimpleMembershipProvider does indeed salt hashes. In fact, the SimpleMembershipProvider makes use of the little known Crypto Helper whose HashPassword method is of most interest in relation to this article:

/* =======================
 * HASHED PASSWORD FORMATS
 * =======================
 * 
 * Version 0:
 * PBKDF2 with HMAC-SHA1, 128-bit salt, 256-bit subkey, 1000 iterations.
 * (See also: SDL crypto guidelines v5.1, Part III)
 * Format: { 0x00, salt, subkey }
 */

public static string HashPassword(string password)
{
    if (password == null)
    {
        throw new ArgumentNullException("password");
    }
    // Produce a version 0 (see comment above) password hash.
    byte[] salt;
    byte[] subkey;
    using (var deriveBytes = new Rfc2898DeriveBytes(password, SaltSize, PBKDF2IterCount))
    {
        salt = deriveBytes.Salt;
        subkey = deriveBytes.GetBytes(PBKDF2SubkeyLength);
    }

    byte[] outputBytes = new byte[1 + SaltSize + PBKDF2SubkeyLength];
    Buffer.BlockCopy(salt, 0, outputBytes, 1, SaltSize);
    Buffer.BlockCopy(subkey, 0, outputBytes, 1 + SaltSize, PBKDF2SubkeyLength);
    return Convert.ToBase64String(outputBytes);
}

This method takes the password, and passes it to the constructor of a Rfc2898DeriveBytes class. The class constructor also needs the size of the salt it should generate (128 bits) and the number of times it should apply the algorithm to the password/salt combination to generate a secure hash. It generates its own cryptographically strong salt and returns that as a byte array along with the hash. The method then constructs a new byte array consisting of (in order) one empty byte (0x00), 16 bytes containing the salt value, and a further 32 bytes containing the hashed salt + password. This is then encoded to a Base64 string for return by the method, and that is the value stored in the Password field.

The Crypto helper also offers another method - VerifyHashedPassword:

// hashedPassword must be of the format of HashWithPassword (salt + Hash(salt+input)
public static bool VerifyHashedPassword(string hashedPassword, string password)
{
    if (hashedPassword == null)
    {
        throw new ArgumentNullException("hashedPassword");
    }
    if (password == null)
    {	
        throw new ArgumentNullException("password");
    }

    byte[] hashedPasswordBytes = Convert.FromBase64String(hashedPassword);

    // Verify a version 0 (see comment above) password hash.

    if (hashedPasswordBytes.Length != (1 + SaltSize + PBKDF2SubkeyLength) || hashedPasswordBytes[0] != 0x00)
    {
         // Wrong length or version header.
         return false;
    }

    byte[] salt = new byte[SaltSize];
    Buffer.BlockCopy(hashedPasswordBytes, 1, salt, 0, SaltSize);
    byte[] storedSubkey = new byte[PBKDF2SubkeyLength];
    Buffer.BlockCopy(hashedPasswordBytes, 1 + SaltSize, storedSubkey, 0, PBKDF2SubkeyLength);

    byte[] generatedSubkey;
    using (var deriveBytes = new Rfc2898DeriveBytes(password, salt, PBKDF2IterCount))
    {
         generatedSubkey = deriveBytes.GetBytes(PBKDF2SubkeyLength);
    }
    return ByteArraysEqual(storedSubkey, generatedSubkey);
}

This method is provided with the password that the user entered on logging in, and the value that is stored in the password field for the particular user. It then reverses part of the work done by the HashPassword method - it converts the Base64 encoded string back to a byte array, and then extracts the hash part of that array, and the salt. It then uses the salt to hash the submitted password, and compares the result to the retrieved hash.

If you only ever need simple (but secure) password storage and verification, you can use the Crypto helper methods without having to configure and use the SimpleMembershipProvider:

@{
    var db = Database.Open("MyDb");
    var username = Request["UserName"];
    var password = Request["Password"];
    var hash = Crypto.HashPassword(password);
    db.Execute("INSERT INTO Users (Username, Password) VALUES (@0, @1)", username, hash);
}

Other interesting methods within the helper offer the ability to generate SHA1 and SHA256 hashes as well as more generic methods that return hashes generated by the algorithm that you specify. There is also a utility method for generating salts.

So why is there an unused column named PasswordSalt in the standard membership table? Who knows, but my guess is that the schema was designed before the Crypto helper, and since the HashPassword method returns one string which represents the encoded and combined salt and the password hash, it was decided not to bother deconstructing that to populate the PasswordSalt field. Either that, or someone simply forgot to remove it.