Storing Connection Strings

Problem

You need to choose the best place to store connection strings that you need in your application to increase maintainability, simplify future modifications, and eliminate the need to recompile the application when it is modified.

Solution

There are several alternatives for storing connection strings, including hard-coding the connection string in your application, storing it in an application configuration file or the Windows Registry, representing it using a Universal Data Link (UDL) file, or in a custom file.

Discussion

A connection string is made up of a semi- colon delimited collection of attribute/value pairs that define how to connect a data source. Although connection strings tend to look similar, the available and required attributes are different depending on the data provider and on the underlying data source. There are a variety of options providing differing degrees of flexibility and security.

Persist Security Info

The Persist Security Info connection string attribute specifies whether the data source can hang on to, or persist, sensitive information such as user authentication credentials. Its value should be kept at the default false . If its value is true , the connection informationincluding the passwordcan be obtained by querying the connection, allowing an untrusted party to have access to sensitive information when a Connection is passed or persisted to a disk. This is an issue only when passing connected objects such as Connection or DataAdapter ; disconnected objects such as DataSet and DataTable do not store information about the original source of their data.

Before a data source object is initialized for the first time, sensitive information can be retrieved from it regardless of the setting of the Persist Security Info property. Avoid passing uninitialized data source objects.

The Persist Security Info connection string attribute is supported by the SQL Server, OLE DB, and Oracle .NET Framework data providers. Although not supported by the ODBC .NET Framework data provider, its behavior is as if Persist Security Info is false and cannot be changed. Check the documentation for other data providers to determine specific implementation details.

Connecting to a database server requires passing credentialsusername and passwordto the server in a connection string. These credentials, together with the data source name , need to be kept private to protect unauthorized access to the data source. There are two approaches for obtaining these credentials:

• Prompting for connection credentials at runtime.
• Storing predetermined connection credentials on the server and using them at runtime to connect to the database server.

Integrated Security

Integrated security is the most secure way to connect to a SQL Server and should be used unless it is impractical to do so. Integrated security uses the identity of the current active user rather than an explicit user ID and password in the connection string to authorize access to the database. Integrated security avoids storing usernames and passwords in connection strings and its use is recommended where possible instead of SQL Server Authentication.

To use integrated security in the connection string, specify the value SSPI for the Integrated Security attribute and do not specify User ID and Password connection string attributes:

Integrated Security=SSPI

See Recipe 1.8 for information about connecting to SQL Server using integrated security from ASP.NET.

Often, it is not practical to prompt for connection credentials because of disadvantages including:

Security

Transferring connection information from the browser to the server can expose connection credentials if they are not encrypted.

Connection pooling

Each user must be recognized separately by the server. This does not allow effective connection pooling and can limit the scalability of the application. For more on connection pooling, see Recipe 1.15.

Single sign-on

It is difficult to integrate with single sign-on strategies, which are becoming increasingly important in enterprise environments (for example, where numerous applications are aggregated into portals).

Server applications

Cannot be used by applications that otherwise have no user interface, such as an XML web service.

There are a number of techniques that you can use to store predetermined connection credentials. These, together with their advantages and drawbacks, are discussed in the following subsections.

Always configure predetermined accounts with the minimum permissions required.

Never use sa or any other administrative account.

Never use blank passwords.

Hardcode in the application

An obvious technique for storing connection strings is hardcoding them into the application. Although this approach results in the best performance, it has poor flexibility; the application needs to be recompiled if the connection string needs to be changed for any reason. Security is poor. The code can be disassembled to expose connection string information. Caching techniques together with external storage techniques eliminate nearly all performance benefits of hardcoding over external storage techniques.

Hardcoding connection string information is not advised; external server-side storage is preferred in nearly all cases because of the increased flexibility, security, and configuration ease. A discussion of available external storage options follows .

Application configuration file

An application configuration file is an XML-based text file that is used to store application-specific settings used at runtime by the application. The naming convention for and deployment location of the file depend on the type of application:

Executable application

The name of the configuration file is the name of the application executable with a .config extensionfor example, myApplication.exe.config . It is located in the same directory as the executable file.

ASP.NET application

A web application can have multiple configuration files all named web.config . Each configuration file supplies configuration settings for its directory and all of its child directories; it also overrides any configuration settings inherited from parent directories.

The machine configuration file machine.config , located in the CONFIG subdirectory of the .NET runtime installationcontains configuration information that applies to the computer. The machine.config file is checked for configuration settings defined in an element before the application configuration file is checked.

It is best to put application settings in the application configuration file both to facilitate deployment and to keep the machine configuration file manageable and secure.

The element of the application file is used to store custom application settings as a collection of key-value pairs. You can store a connection string as shown:

The AppSettings property of the System.Configuration.ConfigurationSettings class is used to retrieve the value for a specific key within the appSettings element; the ConfigurationSettings class cannot be used to write settings to a configuration file.

Application configuration files facilitate deployment because the files are simply installed alongside other application files. One drawback is that application configuration files are not inherently secure since they store information as clear text in a file that is accessible through the file system. Encrypt the connection and other sensitive information within the configuration file and ensure that NTFS file permissions are set to restrict access to the file. Recipe 5.7 shows techniques to encrypt data.

Make sure you name the application configuration file for a Windows Forms application App.config this is the default. At build time, this file is automatically copied into the startup directory by Visual Studio .NET with the name applicationName.exe.config .

If you name the application configuration file applicationName.exe.config within your solution, you will have to copy it to the startup directory each time you modify it and each time you build the solution; the build process deletes it from the startup directory.

Universal data link (UDL) file

The OLE DB .NET data providers supports UDL filenames in its connection string. The UDL file is a resource external to the application that encapsulates connection properties in a separate file. It must be protected using NTFS security to prevent connection information from being exposed or altered . The SQL Server .NET data provider does not support UDL files in its connection string. UDL files are not encrypted; cryptography cannot be used to increase security. NTFS directory and file encryption can secure a UDL file so that even if unauthorized access is gained to the file or the physical disk is stolen, the user ID and password of the user who encrypted the file would still be required to access its contents.

NTFS Encryption

NTFS was enhanced in Windows 2000 with the Encrypted File System (EFS) that provides file- and directory-level encryption. Actually, EFS encrypts only filesdirectories are simply marked so that new files in the directory are encrypted. Encryption and decryption of files is both automatic and transparent for the user who set the encryption.

Encrypted files are visible to any user who can access the system but the contents of the encrypted files can only be viewed by the user who set the encryption. If necessary, standard NT security methods can hide directories and files from view of specific users and user groups.

EFS is a separate mechanism that is used together with the standard security subsystem.

Windows registry

You can store connection strings in the Windows registry as a subkey of HKEY_LOCAL_MACHINESOFTWARE . You can encrypt these settings within the registry subkey and restrict access to the subkey to increase the security of this technique. This technique is easy to use because of programmatic support for registry access in .NET classes Registry and RegistryKey in the Microsoft.Win32 namespace.

Storing connection strings in the registry is usually discouraged because of deployment issues; the registry settings must be deployed with the application, defeating benefits of xcopy deployment. Application code can also be restricted in its access to the registry, further complicating deployment.

Custom file

A custom file is any file that is used to for proprietary storage of application settings that are typically used at runtime. There is generally no particular advantage to using a custom file to store connection information so the technique is not recommended. The approach requires extra coding and forces concurrency and other issues to be explicitly addressed.