Understanding the Three A’s of Security for SQL Server

Do you know what the “Three A’s of Security” are and how they apply to Microsoft SQL Server? Let’s look at them as they are important for managing security properly on a given Microsoft SQL Server.


Authentication is determining who a person/process is.

When a connection is made, it’s important to know what account is making that connection. Authentication is the process of validating who that connection is. There are two types of authentication with respect to users connecting to SQL Server:

  • SQL Server’s build-in logins
  • Windows logins

In the first case, SQL Server has stored logins with passwords and if SQL Server is set to mixed mode authentication, these logins can be used to make a connection to a given SQL Server. More on mixed mode authentication in another post. In the second case, a Windows user can connect to SQL Server. This access can either be granted by to the Windows user directly or via a Windows security group.

Obviously, there are times when SQL Server is able to authenticate a connection but that connection isn’t allowed access to any resources, to include simply logging in to the SQL Server itself. That leads us to the next A.


Authorization is determining what a person/process is allowed to do.

Once SQL Server knows who the connection is, the next question is what should that connection have access to? This is authorization. The initial check is to see if the account in question should even be able to log in to the SQL Server. If it does have that permission (you’ll see it as CONNECT SQL if you look at the permissions), the next question is what other permissions should it have. This can be at the server level and it can be within the individual databases.


Auditing is keeping track of what happens for a person/process.

Auditing is the record keeping involved with tracking various security events. For instance, SQL Server can track both successful and failed logins. Typically, it’s recommended that at least failed logins be tracked. SQL Server can, depending on the edition, audit just about anything you want to track. We’ll cover auditing in more detail in later posts.


Encrypt usernames and passwords stored in files

I was looking at a product recently and came across a rather unpleasant surprise: the install instructions specified that I put the database connection in plaintext in web.config. I’ll explore this particular case and why it’s particularly egregious, but from a security perspective, this shouldn’t happen anymore, regardless of application. We have the tools and the processing power where encrypting data of this sort should be a no-brainer.

The Specific Case

When I say web.config, that should reveal that the application is an ASP.NET web application. It is. And modern ASP.NET apps (anything ASP.NET 2.0+) have the ability to store encrypted keys. The keys are stored encrypted in the registry and the way to get the keys are stored in the web.config file. Therefore, there is no need for anything sensitive to be in plaintext.

The way to do this is with the tool aspnet_regiis.exe. Here are the instructions on MSDN for how to use it in Microsoft.NET Framework 4.0. You can hit the drop down for other versions which will take you back all the way to SQL Server 2005 and 2.0. This feature has been around for about 10 years, as this old document indicates. Since it’s been around that long, there’s no good reason not to do this as a standard course of action for a web application.

But What If the Attacker Gets Admin Rights?

This isn’t to say that you have to use this method. It’s free. It works. However, the point is that credentials should not be left in plaintext. Yes, an attacker who gets administrative access to a server will likely have the resources to be able to decrypt and eventually get the information. However, this is a relatively easy step that makes things harder. It’s a low cost defense. Therefore, it should be implemented.

It Also Means the Username / Password Doesn’t Show Up via Searches

The other thing that encrypting the username and password or even the whole connection string is it means searches of files for this kind of information fail. It’s a trivial thing to run a search and see if you can get some additional nuggets like a username/password combination. Therefore, most smart attackers do these kinds of things. When I pen test a system, I certainly do. Encrypt and the search won’t pick up the information. That’s a good enough reason in and of itself.