We Keep Coding

Best Practice for Connecting ASP.NET to SQL Server

We have an ASP.NET 4.0 Web application that connects to a SQL Server on a separate machine across a LAN. I use a ConnectionString (with SQL Server authentication) stored in my Web.config to do this. Basically, it's a fairly traditional Web-Server-to-SQL strategy.
However, one of our clients is arguing that this strategy is not secure. This client says that we should only connect to the SQL Server through a separate Web Services layer.
I really don't want to rewrite this app just to satisfy this client. What should I tell him? Does any one know how I might best refute this?
Thanks in advance...

Security is always a trade-off. What is the client really afraid of?
Having database credential "in the clear"? I have seen auditors point this out as a potential vulnerability, but really, if someone has compromised your web server they can run arbitrary code against the database, so encrypting database credentials doesn't really buy you much.
Your web app should be using a minimal-rights user to connect to the database, so compromising the web server should only give you the rights to read & update data. How would that change if everything went through a web services layer? Again, there is a very real cost - in complexity, and in performance - by going to a web services layer. Only the client can answer whether or not that cost is worth it.

If this is a web project, you need to change IIS servers running user to a domain user and give permission on sql server to that user.
Than you can use SSPI on your connection string like below.
Like this, you don't need to keep your username or password clearly on web.config.
<configuration>
<system.web>
<identity impersonate="true"/>
</system.web>
and your connectionString
"Integrated Security=SSPI;Initial Catalog=TestDb;Data Source=10.10.10.10"

There are many customers that argue the work of an IT professional, just like there are many people visiting the doctor asking for the medicine instead of what disease they have, because they already know the answer since they read about it on the internet.
I mean, they ask you to build the application and you as an IT professional should know best when your application works as expected. You as a professional should have balls to tell your customer that if he think can get somewhere else better, he should go there or perhaps build the application himself; that's what have done in the past with positive results :)
Regarding security; perhaps for their confidence you can encrypt the web.config and show them, but actually it means nothing; if someone can access the server, they could decode it. On the other hand, someone that want to break in to your database should pass trough a lot of barriers. It's hard to break in, perhaps impossible. Another options is simply blocking connections from outside network the network or ip range or whatever. I think this shouldn't be something to worry about.
There much more and either more realistic concerns to worry about, such as preventing cross site scripting and such common treats.

The client is wrong introducing another tier would not automatically improve security.
In a nutshell use SQL server roles for data access for example the built in data_reader and data_writer roles are a good place to start. Always use the most appropriate least privilege account for the application. If you only need to read data use an account that only has access to read.
Use Windows authentication where possible, if this isn't possible then at least encrypt the connectionstring.
More information on how to do what I've described can be found at http://msdn.microsoft.com/en-us/library/ff650037.aspx#pagpractices0001_dataaccess

One possibility is to encrypt the section in the web.config. So, only user who can access the webserver directly can decrypt this section.
Here is how this works with the help of the iisreg-tool:
http://msdn.microsoft.com/en-us/library/zhhddkxy.aspx

You could enable Encrypted connections in your database and tell the client that the connections are encrypted so fully secure?

To make it more secure, and satisfy your customer you can use Tunneling between your computers.
You setup a server tunneling program where the database exist, and client tunneling programs on the client computer. You connect the one computer to the other via tunneling, and the database connection happens over tunneling.
And everything is exchange high secure (and compressed if the tunneling support it).
http://en.wikipedia.org/wiki/Tunneling_protocol
http://en.wikipedia.org/wiki/HTTP_tunnel
Ps I connect to my server only via tunneling for anything that I do.

I never liked using the web config. The registry is more secure.
Best Practices is to:
Hide the important items of the connection string in the registry
Encrypt the important items in the connection like user names, passwords and server name in the
registry
Access the registry through a class
Build the connection string on the fly and only when needed per page
Error handle every page so the connection string won't show, in case of error
Always close the connections once done. Avoid memory leaks
Always close and reset DataReaders
For additional Security
You can build a separate program to create the connection string and
reference that project as a library inside the solution
If you want to be really secure your client is correct. Send the information to a dll that will communicate with the DB. This is a lot of work
Sources:
From ScottGu # http://msdn.microsoft.com/en-us/library/Aa302406
Scalability:
In regards to scalability: Creating\Editing registry keys across a web farm can be accomplished very easily, with custom admin software.
http://weblogs.asp.net/scottgu/archive/2010/09/08/introducing-the-microsoft-web-farm-framework.aspx
On a final note to all the people who voted me down. Security out of the box is just not enough. Security is an Art and not Science.
Hackers know where the password is stored by default......
ASP.NET 4.0 Fans
Microsoft makes it easy for asp.net 4.0 web sites to deploy registry settings:
http://msdn.microsoft.com/en-us/library/dd394698.aspx

c# protect a database connection string in Settings prevent Decompiling?

Is there anyway to prevent people from using Reflector.net to decompile my .exe c# application? I know there is a tons of post about this but I don't really care if people can see my code the only thing I want to "hide" is my database connection string.
I am currently using "Settings" in my c# to keep the database connection's info.
I wanted to know if using those string in my project's Settings would prevent people from seeing it ?
I am using DotFuscator in visual studio 2008 but I heard it wasn't preventing people from decompiling my program.
I know I could use a Web Services but my server will be on linux so I guess I can't store web services on Linux.

No. Even if you encrypt the connection string in the program code or in a settings file, you will need to decrypt it, and the program must necessarily contain the decryption key somewhere, which means that someone who is interested enough in finding it will find it, no matter how creative you are in hiding it. Why do you need to hide the connection string? If you are afraid that someone who has your program might call the web services directly and trigger unintended actions, you should look into how the web services are structured, what they allow clients to do, and how the authorization works, and make security improvements there instead.

If your program has the connection string in it, users of your program can get it back out. Even if you encrypt it, they can sniff it when your program connects to the DB server.
If you don't want your users to know your DB login credentials, don't give your DB login credentials to the users. That's the only way.
You could do this by instead giving each user their own credentials, and using the permissions system in the DB server to control what they can or can not do.

As others have stated obfuscation is no real protection for a connection string stored in a client application where the user have access to the binaries.
Don't use a direct database connection from your program unless the user is trusted to use the database directly with the same privileges. Have a service (web service, REST-service, etc) in between that you host on your own server. Linux can host services of any of those types I mentioned (use Mono if you want them in .NET on Linux)
In order to expose your database via a web service using Mono or any other language/framework you can host on Linux you would create a web service method for each atomic operation you want to perform against the database.
An additional advantage over letting the client application access the database directly is that when the client application is using a service between itself and the database you are free to change your data store without affecting the client. You can decide to change the database schema in your database or replace the database with a NOSQL solution or even a flat file.
Having a service instead of communicating directly with the database moves the authentication/authorization requirement one step, so now you need to implement it in the service. Fortunately there is rich support for authentication in a web service.

Take a look at this guide on this specific topic from MSDN. Keep in mind, however that this only shifts the security burned. Now you need to manage the security of the key

saving a text file in client side without asking for permission any time

I need to save a text file on the client side possibly without permission. The case is that I need to save this text file in a shared folder in this or in another machine in the lan. This text file is going to be read automatically by the fiscal printer which will print the fiscal invoice. I have a asp .net web application and the server is not on the same lan with the fiscal printer, so I have to write it on the client-side. Any idea how to do this without asking to the user every time for the security issue.
I need a cross browser solution.
I can accept a solution like, the client is asked only one time a the first printing, but not every time he wants to print a bill. Some kind of asking permission to the client for allowing this website, in order to not repeat the permission asking.

Obviously - this would be a major security breach to download files to the user's computer without them knowing. All browsers have precautions in place to prevent this from happening.

No, you can not do this. Saving a file to a computer without permission in a public folder is not allowed.
You can, however, have your Client install your application which will have the ability to read and write where you want.
A common way that Trojan viruses to this is by giving the Client some goofy program to run that displays a fireworks show or something else quite trivial. While the Client is busy wondering what he's looking at, your virus is installing quietly in the background.
Now, you are probably saying to yourself, "But I am not installing a virus." However, there is no way for a Browser to know if your application is a virus or not. That is why it is not allowed and why you can not do it.

The more applicable scenario for me is:
1- Do your work inside your web application.
2- Get the information that you need to print.
3- Send it to another computer directly (or to a hosted web service) and this computer will act as a host for these files.
4- let your server access this shared folder, and print what you want

You could use a cookie, which won't ask permission. Of course that would only work when cookies are enabled and can store limited amounts of data.

Can't get my .Net Console App to run on a Server "The Publisher Could Not Be Verified"

I've written a simple C# console app that gets executed by SQL Server Integration Services as part of a datawarehouse load (on Windows Server 2003).
Unfortunately the step that calls the .exe is failing because a Open File security warning dialog "The Publisher Could Not Be Verified. Are you sure you want to run this software" is shown.
I think I need to sign my console app, but this seems a little overkill for an internal application. Can I sign it without having to get an externally recognised certificate?
Thanks.

Yes, signing is easy, Just look up Project|Options|Signing. The combobox will let you pick a key-file or create one. It is always a good idea.
But signing may not be enough, the server may impose stricter rules so that you will need a certificate. I'm not 100% sure but you may be abkle to create your own and tell the server to trust it.
But try the signing first.

C# mysql connection practices

If a C# application connects to a mysql server from a client, how do I store the mysql username/password for the connection? If I have it in a config file or embedded in the source it can be found by reverse engineering. It is not possible to give all users a MySql password.
Also, I have a log in for the application. How do I enforce that the user goes through the login process and does not just reverse engineer and comment out the C# code verifying the log in?
Is there anyway manage these connections between MySql and a client side application or must there be a third program on the server side interacting with the database locally to be secure?

Perhaps you can have a two-stage system: Have a SQL account whose only permission is to execute a stored procedure taking the user's username and password and giving them the credentials to use for the real account. When the user logs in, you connect using the restricted account, get the credentials for the real account, and then do your work using that account. You can change the SQL password fairly frequently, too. If you suspect a user of foul play, have the procedure return them a different set of credentials, and track those credentials.

For Winform clients that connect directly to the db this is an age old (10 years or so?) question that may never be solved.
The problem is that no matter how you encrypt or obfuscate the connection string there will always be a point in time at which the string will be represented as plain text in the client computer's memory and therefore hackable.
You have been recommended options by other SOers, i just thought i'd point out what you're trying to work around.

Here is the problem. You are trusting the end user with the binaries which will call MySQL queries. This means, beyond any shadow of a doubt, that a clever user could "take control" and run queries directly.
There are things you can do to improve the situation. It sounds like you are on a LAN. Why can't you give each user their own database user? That means that the authentication is (a) taken care of for you, and (b) you can use "real" MySQL permissions to limit what harm they can do. Also, you could use stored procedures and give them only access to the procs, really limiting what they can do.
You could also consider re-writing as a web-application where you process everything on the server out of their reach.
However, is there really a problem here, or are you just being theoretical?