User Authentication, Password Handling, and Doing It Safely

[NanoBob]

Introduction

Properly handling your user's credentials (username and password) is very important, this guide gives detailed information and code samples on how to (properly) implement an account "system". This guide assumes you are not using MTA's built in accounts. 

Disclaimer: Any code shown in this tutorial is purely for illustrative purposes, and is not finished code, you should use it as a guideline, not a solution.

Content

The following topics will be discussed in this tutorial:

• How to hash and salt passwords (register)
• How to validate a hashed password (login)
• How to add "remember-me" functionality
• How to offer password recovery
• How to migrate from an older hashing algorithm, to a newer one
• Using a password policy
• (Extra) How to handle database leaks
• (Extra) What even is hashing and salting?

For the purpose of this tutorial we expect a database structure which is somewhat similar to this:

How to hash and salt passwords
When you have a user register on your service, that user expects of you to keep their password safe. Whilst it is generally bad practice to use the same password for multiple services there are many users that still do so. Because of this it's crucial that you save the user's passwords in a way that an attacker will be unable to find out the original password the user typed. This includes if they have full access to your database. In order to do this we do what is called "Password hashing"

When a user registers, your server receives the user's intended username, (email) and password. Before you save that password to the database you have to hash and salt this, luckily MTA has a function that takes care of this. If you wish to know more about what exactly it does, there's more information at the end of this tutorial. 

In order to hash this password you use the passwordHash function. This function is relatively slow (by design), so it is highly recommended you pass a callback to this function, so your entire script doesn't wait for it to complete.

https://wiki.multitheftauto.com/wiki/PasswordHash

local mysqlHandle -- we're assuming this value is set somewhere else in code

function register(username, email, password)
    local player = client

    passwordHash(password, "bcrypt", {}, function(hashedPassword)
        -- callback function for hashing the password
        local handle = dbQuery(function(handle) 
            -- callback function for storing the user in the database
            if (handle) then
                triggerClientEvent(player, "registrationSuccess") -- inform the user that registration was successful
            else
                triggerClientEvent(player, "registrationFailed")
            end
        end,mysqlHandle, "INSERT INTO users (email, username, password) VALUES (?, ?, ?)", email, username, hashedPassword)

    end)

end
addEvent("passwordTutorial:register", true)
addEventHandler("passwordTutorial:register", getRootElement(), register)

 

How to validate a hashed password
Once you've saved the hashed password to your database you need to do a little bit of additional work when authenticating the user. Luckily MTA offers a passwordVerify() function, which is the counterpart of the previously discussed passwordHash(). What this function does it basically hashes the password in the same way, resulting in the same output hash.

https://wiki.multitheftauto.com/wiki/passwordVerify

In order to get the account the user is trying to log in to you have to do a query for an account which has the user submitted username, and of which the password matches through passwordVerify. PasswordVerify is also a relatively slow function, thus you should use a callback.

 

function login(username, password)
    local player = client

    dbQuery(function (handle)
        -- callback for the query selecting the user by username
        local results = dbPoll(handle, -1)
        if (#results == 0) then
            triggerClientEvent(player, "loginFailed")
            return
        end

        passwordVerify(password, results[1].password, {}, function(matches)
            -- callback function for the password verify
            if (matches) then
                -- Do anything you wish with the database result to log the player in with the rest of your scripts
                triggerClientEvent(player, "loginSuccess")
            else
                triggerClientEvent(player, "loginFailed")
            end
        end)

    end, mysqlHandle, "SELECT * FROM users WHERE username = ?", username)
end
addEvent("passwordTutorial:login", true)
addEventHandler("passwordTutorial:login", getRootElement(), login)

 

How to add "remember me" functionality
When users on your server log in, they often do not want to have to enter their username and password every time they want to log in. In order to satisfy this need you can implement a "remember me" function. What I've seen happen in the past, is people would store the user's password (encrypted) on the client. This is NOT safe, and should never be done!

In order to properly use remember me functionality what you would do is upon login in, generate a random string. The longer the better. This random string is what we call an access token. You would then allow the user to log in with such an access token, preferably only once generating a new access token each time one is used.

To implement this you would generate that token every time the user logs in, whilst they have "remember me" enabled. You will have to save this token in your database alongside your user.  For extra security you could also store the user's serial alongside the access token, you can then validate that the access token is being used from the same device.

https://wiki.multitheftauto.com/wiki/Filepath

function login(username, password)
    -- This code should be put in the callback to the dbQuery function, but to keep the example clean that's not shown here
    if (rememberMe) then
        local token = generateRandomToken()
        dbQuery(mysqlHandle, "INSERT INTO access_tokens (user_id, token) VALUES (?, ?)", results[1].id, token)
        triggerClientEvent(player, "loginSuccess", token)
    end
end

function rememberMeLogin(username, accessToken)
    -- this function handles a user's login attempt 
    dbQuery(function(handle)
        local result = dbPoll(handle, -1)
        if (#result == 0) then
            triggerClientEvent(player, "loginFailed")
        else
            -- Do anything you wish with the database result to log the player in with the rest of your scripts
            triggerClientEvent(player, "loginSuccess")
        end
    end,mysqlHandle, "SELECT users.* FROM access_tokens JOIN users ON users.id = access_tokens.user_id WHERE users.username = ?", username)
end
addEvent("passwordTutorial:loginRememberMe", true)
addEventHandler("passwordTutorial:loginRememberMe", getRootElement(), login)

How to offer password recovery
Offering password recovery requires a little bit more than just your MTA server. Generally password recovery is done with emails. So you would need access to an email server / service which you can use to send an email from an HTTP request. (Like you can do with fetchRemote()). 

When a user requests a password reset, have them enter the email you registered with. You then fetch a user from the database with this email address. You would then store a password recovery token for this user. This token, just like the remember me token, is a random string. 

Ideally, you would send the user a link with a password reset form that goes to a webpage where the user can reset their password. You could do this with an external service, like a webserver. Or you could use MTA's Resource web access for it, but if you do make sure you handle permissions properly for anything else that uses this. However another option would be to have the user copy paste the generated token from the email into you server's login window. 

Which of the two solutions you pick is up to you, my personal preference goes to the one with the link in the email. But in either case the server side logic is the same.

When the user attempts to perform password recovery, verify that the token they give you belongs to a user, and then change the password to the newly requested password. Make sure you hash this password the same way you do in your login.

function requestPasswordRecovery(email)
    dbQuery(function (handle))
        local result = dbPoll(handle, -1)
        if (#result == 0) then
            triggerClientEvent(player, "passwordTutorial:passwordRecoveryRequestFailed")
        else
            local token = generateRandomToken()
            dbExec(mysqlHandle, "UPDATE user_data SET recovery_token = ?", token)
            -- mail the token to the user, mail implementation depends on the mail server/service you use
            triggerClientEvent(player, "passwordTutorial:passwordRecoveryRequestSuccess")
        end
    end, mysqlHandle, "SELECT * FROM users WHERE email = ?", email)
end

function recoverPassword(recoveryToken, password)
    dbQuery(function (handle)
        local result = dbPoll(handle, -1)
        if (#result == 0) then
            -- this is only valid if you have the user request password recovery from ingame
            triggerClientEvent(player, "passwordTutorial:passwordRecoveryFailed")
        else

            passwordHash(password, "bcrypt", {}, function(hashedPassword)
                -- callback function for hashing the password
                local handle = dbExec(function(handle) 
                    -- callback function for storing the new password in the database
                    if (handle) then
                        -- this is only valid if you have the user request password recovery from ingame
                        triggerClientEvent(player, "passwordTutorial:passwordRecoverySuccess") -- inform the user that registration was successful
                    else
                        -- this is only valid if you have the user request password recovery from ingame
                        triggerClientEvent(player, "passwordTutorial:passwordRecoveryFailed")
                    end
                end,mysqlHandle, "UPDATE user_data SET password = ? WHERE recovery_token = ?", username, recoveryToken)
        
            end)

        end
    end, "SELECT * FROM users WHERE recovery_token = ?", recoveryToken)
end

Besides changing the password, it's important you also delete any access tokens that user might have if you're using remember me functionality.
It is also good practice to make recovery tokens expiry after a certain amount of times, and not allow a recovery token to be created whilst one is already in progress. This prevents a user from sending a large number of emails from your service.

How to migrate from an older hashing algorithm, to a newer one
Maybe after reading this topic you realise that your password security is not what it should be. So you want to change your old password hashing / validation logic to the ones explained in this topic. And due to the nature that hashes can not be "unhashed", you can't simply migrate your passwords over. So in order to migrate the passwords what you have to do is when a user logs in, first validate their password with the old hashing algorithm. If this matches, then hash (and salt) it with your new hashing algorithm and save it in your database. Make sure to delete the old password otherwise your password security is not any better than before.

Using a password policy
Passwords policies are important to prevent your users from picking a password that is too easily cracked / brute forced. Many password policies come in the form of "Must have at least one capital letter, one digit and one number". But that discards that fact that the best way to make your password more difficult to crack, is making your password longer. So in the code snippet below is a function that measures the 'search space' of a password. The search space of a password is the amount of possible passwords there are with a certain combination of characters.

In order to use this, you would have to set a minimum password search space when a user registers for an account. This minimum is up for you to set, but be reasonable, you shouldn't expect a user's password to be impossible to remember / create. I recommend playing with the function a bit to see what values you get out of it, and pick something you believe is sensible.

function getPasswordSearchSpace(password)
    local lowerCase = password:find("%l") and 26 or 0
    local upperCase = password:find("%u") and 26 or 0
    local digits = password:find("%d") and 10 or 0
    local symbols = password:find("%W") and 32 or 0
    local length = password:len()

    return (lowerCase + upperCase + digits + symbols) ^ length
end

-- The below function calls are to indicate the difference in search space for a set of passwords
print(getPasswordSearchSpace("a"))
print(getPasswordSearchSpace("abc"))
print(getPasswordSearchSpace("Abc"))
print(getPasswordSearchSpace("Ab!"))
print(getPasswordSearchSpace("Ab!0"))

print(getPasswordSearchSpace("Mu#9A0h."))
print(getPasswordSearchSpace("This is a demonstration of how easy an incredibly strong password is to remember"))

 

How to handle database leaks
If you have reason to believe that your database has been leaked or otherwise compromised, it is important that your first course of action is removing any access tokens stored in your database. Once you have done that you have to inform your users. Whilst when properly hashed and salted it's extremely difficult / time consuming to find out a user's password it is still a possibility. So you should inform your users of the breach, tell them that their passwords were properly hashed, and they do not need to fear for their passwords immediately. However you should suggest to your users that they change their password either way, just in case.

What even is hashing and salting?
Hashing has been brought up several times in this tutorial, whilst you do not need to know what it is / does, you might be interested in knowing regardless. I won't be going too far in depth as I simply do not have the knowledge, but the basic idea of hashing is this:

When you hash anything, you turn it into a string of characters (or other value) that has no relation to the original input, other than when you hash the original input again, it will always generate the same hash.
For example, when you hash the string 'banana' using the sha512 hashing algorithm, it will always yield the output: "F8E3183D38E6C51889582CB260AB825252F395B4AC8FB0E6B13E9A71F7C10A80D5301E4A949F2783CB0C20205F1D850F87045F4420AD2271C8FD5F0CD8944BE3"

Now hashing can not be reverted, you can not "unhash" a hash, so in order to verify someone's password you hash it again, and see if the two hashes are the exact same. Now this is great, passwords are safely stored. However there is still more to do, salting.

Salting is adding some random data to your password prior to hashing it. This prevents when two users (on the same service, or on others) have the same password, that their hashes are also the same. Meaning if one password is compromised, the other password is not. It is important that a salt is random for every user in your application, not one salt for your entire application.

Now you might think we didn't do any salting in the code / tutorial above. This is not true, we just didn't do it ourselves. MTA's passwordHash function actually hashes the passwords and salts it, this salt is then stored in the output hash it self, just before the actual password hash. In the case of bcrypt it actually stores a little bit more info in the resulting hash, but you need not worry about that.

Edited December 27, 2019 by NanoBob