Why does Active Directory validate last password?

I am working on a simple solution to update a user's password in Active Directory.

I can successfully update the users password. Updating the password works fine. Lets say the user has updated the password from MyPass1 to MyPass2

Now when I run my custom code to validate users credential using:

using(PrincipalContext pc = new PrincipalContext(ContextType.Domain, "TheDomain"))
{
    // validate the credentials
    bool isValid = pc.ValidateCredentials("myuser", "MyPass2");
}

//returns true - which is good

Now when I enter some wrong password it validates very nicely:

using(PrincipalContext pc = new PrincipalContext(ContextType.Domain, "TheDomain"))
{
    // validate the credentials
    bool isValid = pc.ValidateCredentials("myuser", "wrongPass");
}

//returns false - which is good

Now for some odd reasons, it validates the previous last password which was MyPass1 remember?

using(PrincipalContext pc = new PrincipalContext(ContextType.Domain, "TheDomain"))
{
    // validate the credentials
    bool isValid = pc.ValidateCredentials("myuser", "MyPass1");
}

//returns true - but why? we have updated password to Mypass2

I got this code from:

Validate a username and password against Active Directory?

Is it something to do with last password expiry or is this how the validation supposed to work?

Answers


The reason why you are seeing this has to do with special behavior specific to NTLM network authentication.

Calling the ValidateCredentials method on a PrincipalContext instance results in a secure LDAP connection being made, followed by a bind operation being performed on that connection using a ldap_bind_s function call.

The authentication method used when calling ValidateCredentials is AuthType.Negotiate. Using this results in the bind operation being attempted using Kerberos, which (being not NTLM of course) will not exhibit the special behavior described above. However, the bind attempt using Kerberos will fail (the password being wrong and all), which will result in another attempt being made, this time using NTLM.

You have two ways to approach this:

  1. Follow the instructions in the Microsoft KB article I linked to shorten or eliminate the lifetime period of an old password using the OldPasswordAllowedPeriod registry value. Probably not the most ideal solution.
  2. Don't use PrincipleContext class to validate credentials. Now that you know (roughly) how ValidateCredentials works, it shouldn't be too difficult for you to do the process manually. What you'll want to do is create a new LDAP connection (LdapConnection), set its network credentials, set the AuthType explicitly to AuthType.Kerberos, and then call Bind(). You'll get an exception if the credentials are bad.

The following code shows how you can perform credential validation using only Kerberos. The authentication method at use will not fall back to NTLM in the event of failure.

private const int ERROR_LOGON_FAILURE = 0x31;

private bool ValidateCredentials(string username, string password, string domain)
{
  NetworkCredential credentials
    = new NetworkCredential(username, password, domain);

  LdapDirectoryIdentifier id = new LdapDirectoryIdentifier(domain);

  using (LdapConnection connection = new LdapConnection(id, credentials, AuthType.Kerberos))
  {
    connection.SessionOptions.Sealing = true;
    connection.SessionOptions.Signing = true;

    try
    {
      connection.Bind();
    }
    catch (LdapException lEx)
    {
      if (ERROR_LOGON_FAILURE == lEx.ErrorCode)
      {
        return false;
      }
      throw;
    }
  }
  return true;
}

I try to never use exceptions to handle the flow control of my code; however, in this particular instance, the only way to test credentials on a LDAP connection seems to be to attempt a Bind operation, which will throw an exception if the credentials are bad. PrincipalContext takes the same approach.


Depending on the context of how you run this, it may have to do with something called "cached credentials."


I've found a way to validate the user's current credentials only. It leverages the fact that ChangePassword does not use cached credentials. By attempting to change the password to it's current value, which first validates the password, we can determine if the password is incorrect or there is a policy problem (can't reuse the same password twice).

Note: this will probably only work if your policy has a history requirement of at least not allowing to repeat the most recent password.

        var isPasswordValid = PrincipalContext.ValidateCredentials(
            userName,
            password);

        // use ChangePassword to test credentials as it doesn't use caching, unlike ValidateCredentials
        if (isPasswordValid)
        {
            try
            {
                user.ChangePassword(password, password);
            }
            catch (PasswordException ex)
            {
                if (ex.InnerException != null && ex.InnerException.HResult == -2147024810)
                {
                    // Password is wrong - must be using a cached password
                    isPasswordValid = false;
                }
                else
                {
                    // Password policy problem - this is expected, as we can't change a password to itself for history reasons    
                }
            }
            catch (Exception)
            {
                // ignored, we only want to check wrong password. Other AD related exceptions should occure in ValidateCredentials
            }
        }

Need Your Help

xUnit : Assert two List<T> are equal?

xunit.net

I'm new to TDD and xUnit so I want to test my method that looks something like:

Managing large binary files with Git

git version-control large-files binaryfiles

I am looking for opinions of how to handle large binary files on which my source code (web application) is dependent. We are currently discussing several alternatives: