Isolated storage

For desktop apps, isolated storage is a data storage mechanism that provides isolation and safety by defining standardized ways of associating code with saved data. Standardization provides other benefits as well. Administrators can use tools designed to manipulate isolated storage to configure file storage space, set security policies, and delete unused data. With isolated storage, your code no longer needs unique paths to specify safe locations in the file system, and data is protected from other applications that only have isolated storage access. Hard-coded information that indicates where an application's storage area is located is unnecessary.

Important

Isolated storage is not available for Windows 8.x Store apps. Instead, use the application data classes in the Windows.Storage namespaces included in the Windows Runtime API to store local data and files. For more information, see Application data in the Windows Dev Center.

Data Compartments and Stores

When an application stores data in a file, the file name and storage location must be carefully chosen to minimize the possibility that the storage location will be known to another application and, therefore, vulnerable to corruption. Without a standard system in place to manage these problems, improvising techniques that minimize storage conflicts can be complex, and the results can be unreliable.

With isolated storage, data is always isolated by user and by assembly. Credentials such as the origin or the strong name of the assembly determine assembly identity. Data can also be isolated by application domain, using similar credentials.

When you use isolated storage, your application saves data to a unique data compartment that is associated with some aspect of the code's identity, such as its publisher or signature. The data compartment is an abstraction, not a specific storage location; it consists of one or more isolated storage files, called stores, which contain the actual directory locations where data is stored. For example, an application might have a data compartment associated with it, and a directory in the file system would implement the store that actually preserves the data for that application. The data saved in the store can be any kind of data, from user preference information to application state. For the developer, the location of the data compartment is transparent. Stores usually reside on the client, but a server application could use isolated stores to store information by impersonating the user on whose behalf it is functioning. Isolated storage can also store information on a server with a user's roaming profile so that the information will travel with the roaming user.

Quotas for Isolated Storage

A quota is a limit on the amount of isolated storage that can be used. The quota includes bytes of file space as well as the overhead associated with the directory and other information in the store. Isolated storage uses permission quotas, which are storage limits that are set by using IsolatedStoragePermission objects. If you try to write data that exceeds the quota, an IsolatedStorageException exception is thrown. Security policy, which can be modified using the .NET Framework Configuration Tool (Mscorcfg.msc), determines which permissions are granted to code. Code that has been granted IsolatedStoragePermission is restricted to using no more storage than the UserQuota property allows. However, because code can bypass permission quotas by presenting different user identities, permission quotas serve as guidelines for how code should behave rather than as a firm limit on code behavior.

Quotas are not enforced on roaming stores. Because of this, a slightly higher level of permission is required for code to use them. The enumeration values AssemblyIsolationByRoamingUser and DomainIsolationByRoamingUser specify a permission to use isolated storage for a roaming user.

Secure Access

Using isolated storage enables partially trusted applications to store data in a manner that is controlled by the computer's security policy. This is especially useful for downloaded components that a user might want to run cautiously. Security policy rarely grants this kind of code permission when you access the file system by using standard I/O mechanisms. However, by default, code running from the local computer, a local network, or the Internet is granted the right to use isolated storage.

Administrators can limit how much isolated storage an application or a user has available, based on an appropriate trust level. In addition, administrators can remove a user's persisted data completely. To create or access isolated storage, code must be granted the appropriate IsolatedStorageFilePermission permission.

To access isolated storage, code must have all necessary native platform operating system rights. The access control lists (ACLs) that control which users have the rights to use the file system must be satisfied. .NET applications already have operating system rights to access isolated storage unless they perform (platform-specific) impersonation. In this case, the application is responsible for ensuring that the impersonated user identity has the proper operating system rights to access isolated storage. This access provides a convenient way for code that is run or downloaded from the web to read and write to a storage area related to a particular user.

To control access to isolated storage, the common language runtime uses IsolatedStorageFilePermission objects. Each object has properties that specify the following values:

  • Allowed usage, which indicates the type of access that is allowed. The values are members of the IsolatedStorageContainment enumeration. For more information about these values, see the table in the next section.

  • Storage quota, as discussed in the preceding section.

The runtime demands IsolatedStorageFilePermission permission when code first attempts to open a store. It decides whether to grant this permission, based on how much the code is trusted. If the permission is granted, the allowed usage and storage quota values are determined by security policy and by the code's request for IsolatedStorageFilePermission. Security policy is set by using the .NET Framework Configuration Tool (Mscorcfg.msc). All callers in the call stack are checked to ensure that each caller has at least the appropriate allowed usage. The runtime also checks the quota imposed on the code that opened or created the store in which the file is to be saved. If these conditions are satisfied, permission is granted. The quota is checked again every time a file is written to the store.

Application code is not required to request permission because the common language runtime will grant whatever IsolatedStorageFilePermission is appropriate based on security policy. However, there are good reasons to request specific permissions that your application needs, including IsolatedStorageFilePermission.

Allowed Usage and Security Risks

The allowed usage specified by IsolatedStorageFilePermission determines the degree to which code will be allowed to create and use isolated storage. The following table shows how the allowed usage specified in the permission corresponds to types of isolation and summarizes the security risks associated with each allowed usage.

Allowed usage Isolation types Security impact
None No isolated storage use is allowed. There is no security impact.
DomainIsolationByUser Isolation by user, domain, and assembly. Each assembly has a separate substore within the domain. Stores that use this permission are also implicitly isolated by computer. This permission level leaves resources open to unauthorized overuse, although enforced quotas make it more difficult. This is called a denial of service attack.
DomainIsolationByRoamingUser Same as DomainIsolationByUser, but store is saved to a location that will roam if roaming user profiles are enabled and quotas are not enforced. Because quotas must be disabled, storage resources are more vulnerable to a denial of service attack.
AssemblyIsolationByUser Isolation by user and assembly. Stores that use this permission are also implicitly isolated by computer. Quotas are enforced at this level to help prevent a denial of service attack. The same assembly in another domain can access this store, opening the possibility that information could be leaked between applications.
AssemblyIsolationByRoamingUser Same as AssemblyIsolationByUser, but store is saved to a location that will roam if roaming user profiles are enabled and quotas are not enforced. Same as in AssemblyIsolationByUser, but without quotas, the risk of a denial of service attack increases.
AdministerIsolatedStorageByUser Isolation by user. Typically, only administrative or debugging tools use this level of permission. Access with this permission allows code to view or delete any of a user's isolated storage files or directories (regardless of assembly isolation). Risks include, but are not limited to, leaking information and data loss.
UnrestrictedIsolatedStorage Isolation by all users, domains, and assemblies. Typically, only administrative or debugging tools use this level of permission. This permission creates the potential for a total compromise of all isolated stores for all users.

Safety of isolated storage components with regard to untrusted data

This section applies to the following frameworks:

  • .NET Framework (all versions)
  • .NET Core 2.1+
  • .NET 5.0+

.NET Framework and .NET Core offer isolated storage as a mechanism to persist data for a user, an application, or a component. This is a legacy component primarily designed for now-deprecated Code Access Security scenarios.

Various isolated storage APIs and tools can be used to read data across trust boundaries. For example, reading data from a machine-wide scope can aggregate data from other, possibly less-trusted user accounts on the machine. Components or applications that read from machine-wide isolated storage scopes should be aware of the consequences of reading this data.

Security-sensitive APIs that can read from the machine-wide scope

Components or applications that call any of the following APIs read from the machine-wide scope:

The isolated storage tool storeadm.exe is impacted if called with the /machine switch, as shown in the following code:

storeadm.exe /machine [any-other-switches]

The isolated storage tool is provided as part of Visual Studio and the .NET Framework SDK.

If the application doesn't involve calls to the preceding APIs, or if the workflow doesn't involve calling storeadm.exe in this manner, this document doesn't apply.

Impact in multi-user environments

As mentioned previously, the security impact from these APIs results from data written from one trust environment is read from a different trust environment. Isolated storage generally uses one of three locations to read and write data:

  1. %LOCALAPPDATA%\IsolatedStorage\: For example, C:\Users\<username>\AppData\Local\IsolatedStorage\, for User scope.
  2. %APPDATA%\IsolatedStorage\: For example, C:\Users\<username>\AppData\Roaming\IsolatedStorage\, for User|Roaming scope.
  3. %PROGRAMDATA%\IsolatedStorage\: For example, C:\ProgramData\IsolatedStorage\, for Machine scope.

The first two locations are isolated per-user. Windows ensures that different user accounts on the same machine cannot access each other's user profile folders. Two different user accounts who use the User or User|Roaming stores will not see each other's data and cannot interfere with each other's data.

The third location is shared across all user accounts on the machine. Different accounts can read from and write to this location, and they're able to see each other's data.

The preceding paths may differ based on the version of Windows in use.

Now consider a multi-user system with two registered users Mallory and Bob. Mallory has the ability to access her user profile directory C:\Users\Mallory\, and she can access the shared machine-wide storage location C:\ProgramData\IsolatedStorage\. She cannot access Bob's user profile directory C:\Users\Bob\.

If Mallory wishes to attack Bob, she might write data to the machine-wide storage location, then attempt to influence Bob into reading from the machine-wide store. When Bob runs an app that reads from this store, that app will operate on the data Mallory placed there, but from within the context of Bob's user account. The remainder of this document contemplates various attack vectors and what steps apps can do to minimize their risk to these attacks.

Note

In order for such an attack to take place, Mallory requires:

  • A user account on the machine.
  • The ability to place a file into a known location on the file system.
  • Knowledge that Bob will at some point run an app that attempts to read this data.

These are not threat vectors that apply to standard single-user desktop environments like home PCs or single-employee enterprise workstations.

Elevation of privilege

An elevation of privilege attack occurs when Bob's app reads Mallory's file and automatically tries to take some action based on the contents of that payload. Consider an app that reads the contents of a startup script from the machine-wide store and passes those contents to Process.Start. If Mallory can place a malicious script inside the machine-wide store, when Bob launches his app:

  • His app parses and launches Mallory's malicious script under the context of Bob's user profile.
  • Mallory gains access to Bob's account on the local machine.

Denial of service

A denial of service attack occurs when Bob's app reads Mallory's file and crashes or otherwise stops functioning correctly. Consider again the app mentioned previously, which attempts to parse a startup script from the machine-wide store. If Mallory can place a file with malformed contents inside the machine-wide store, she might:

  • Cause Bob's app to throw an exception early in the startup path.
  • Prevent the app from launching successfully because of the exception.

She has then denied Bob the ability to launch the app under his own user account.

Information disclosure

An information disclosure attack occurs when Mallory can trick Bob into disclosing the contents of a file that Mallory does not normally have access to. Consider that Bob has a secret file C:\Users\Bob\secret.txt that Mallory wants to read. She knows the path to this file, but she cannot read it because Windows forbids her from gaining access to Bob's user profile directory.

Instead, Mallory places a hard link into the machine-wide store. This is a special kind of file that itself does not contain any contents, rather, it points to another file on disk. Attempting to read the hard link file will instead read the contents of the file targeted by the link. After creating the hard link, Mallory still cannot read the file contents because she does not have access to the target (C:\Users\Bob\secret.txt) of the link. However, Bob does have access to this file.

When Bob's app reads from the machine-wide store, it now inadvertently reads the contents of his secret.txt file, just as if the file itself had been present in the machine-wide store. When Bob's app exits, if it attempts to resave the file to the machine-wide store, it will end up placing an actual copy of the file in the *C:\ProgramData\IsolatedStorage* directory. Since this directory is readable by any user on the machine, Mallory can now read the contents of the file.

Best practices to defend against these attacks

Important: If your environment has multiple mutually untrusted users, do not call the API IsolatedStorageFile.GetEnumerator(IsolatedStorageScope.Machine) or invoke the tool storeadm.exe /machine /list. Both of these assume that they're operating on trusted data. If an attacker can seed a malicious payload in the machine-wide store, that payload can lead to an elevation of privilege attack under the context of the user who runs these commands.

If operating in a multi-user environment, reconsider use of isolated storage features that target the Machine scope. If an app must read data from a machine-wide location, prefer to read the data from a location that's writable only by admin accounts. The %PROGRAMFILES% directory and the HKLM registry hive are examples of locations that are writable by only administrators and readable by everyone. Data read from those locations is therefore considered trustworthy.

If an app must use the Machine scope in a multi-user environment, validate the contents of any file that you read from the machine-wide store. If the app deserializing object graphs from these files, consider using safer serializers like XmlSerializer instead of dangerous serializers like BinaryFormatter or NetDataContractSerializer. Use caution with deeply nested object graphs or object graphs that perform resource allocation based on the file contents.

Isolated Storage Locations

Sometimes it is helpful to verify a change to isolated storage by using the file system of the operating system. You might also want to know the location of isolated storage files. This location is different depending on the operating system. The following table shows the root locations where isolated storage is created on a few common operating systems. Look for Microsoft\IsolatedStorage directories under this root location. You must change folder settings to show hidden files and folders in order to see isolated storage in the file system.

Operating system Location in file system
Windows 2000, Windows XP, Windows Server 2003 (upgrade from Windows NT 4.0) Roaming-enabled stores =

<SYSTEMROOT>\Profiles\<user>\Application Data

Nonroaming stores =

<SYSTEMROOT>\Profiles\<user>\Local Settings\Application Data
Windows 2000 - clean installation (and upgrades from Windows 98 and Windows NT 3.51) Roaming-enabled stores =

<SYSTEMDRIVE>\Documents and Settings\<user>\Application Data

Nonroaming stores =

<SYSTEMDRIVE>\Documents and Settings\<user>\Local Settings\Application Data
Windows XP, Windows Server 2003 - clean installation (and upgrades from Windows 2000 and Windows 98) Roaming-enabled stores =

<SYSTEMDRIVE>\Documents and Settings\<user>\Application Data

Nonroaming stores =

<SYSTEMDRIVE>\Documents and Settings\<user>\Local Settings\Application Data
Windows 8, Windows 7, Windows Server 2008, Windows Vista Roaming-enabled stores =

<SYSTEMDRIVE>\Users\<user>\AppData\Roaming

Nonroaming stores =

<SYSTEMDRIVE>\Users\<user>\AppData\Local

Creating, Enumerating, and Deleting Isolated Storage

.NET provides three classes in the System.IO.IsolatedStorage namespace to help you perform tasks that involve isolated storage:

The isolated storage classes enable you to create, enumerate, and delete isolated storage. The methods for performing these tasks are available through the IsolatedStorageFile object. Some operations require you to have the IsolatedStorageFilePermission permission that represents the right to administer isolated storage; you might also need to have operating system rights to access the file or directory.

For a series of examples that demonstrate common isolated storage tasks, see the how-to topics listed in Related Topics.

Scenarios for Isolated Storage

Isolated storage is useful in many situations, including these four scenarios:

  • Downloaded controls. Managed code controls downloaded from the Internet are not allowed to write to the hard drive through normal I/O classes, but they can use isolated storage to persist users' settings and application states.

  • Shared component storage. Components that are shared between applications can use isolated storage to provide controlled access to data stores.

  • Server storage. Server applications can use isolated storage to provide individual stores for a large number of users making requests to the application. Because isolated storage is always segregated by user, the server must impersonate the user making the request. In this case, data is isolated based on the identity of the principal, which is the same identity the application uses to distinguish between its users.

  • Roaming. Applications can also use isolated storage with roaming user profiles. This allows a user's isolated stores to roam with the profile.

Do not use isolated storage in the following situations:

  • To store high-value secrets, such as unencrypted keys or passwords, because isolated storage is not protected from highly trusted code, from unmanaged code, or from trusted users of the computer.

  • To store code.

  • To store configuration and deployment settings, which administrators control. (User preferences are not considered to be configuration settings because administrators do not control them.)

Many applications use a database to store and isolate data, in which case one or more rows in a database might represent storage for a specific user. You might choose to use isolated storage instead of a database when the number of users is small, when the overhead of using a database is significant, or when no database facility exists. Also, when the application requires storage that is more flexible and complex than what a row in a database provides, isolated storage can provide a viable alternative.

Title Description
Types of Isolation Describes the different types of isolation.
How to: Obtain Stores for Isolated Storage Provides an example of using the IsolatedStorageFile class to obtain a store isolated by user and assembly.
How to: Enumerate Stores for Isolated Storage Shows how to use the IsolatedStorageFile.GetEnumerator method to calculate the size of all isolated storage for the user.
How to: Delete Stores in Isolated Storage Shows how to use the IsolatedStorageFile.Remove method in two different ways to delete isolated stores.
How to: Anticipate Out-of-Space Conditions with Isolated Storage Shows how to measure the remaining space in an isolated store.
How to: Create Files and Directories in Isolated Storage Provides some examples of creating files and directories in an isolated store.
How to: Find Existing Files and Directories in Isolated Storage Demonstrates how to read the directory structure and files in isolated storage.
How to: Read and Write to Files in Isolated Storage Provides an example of writing a string to an isolated storage file and reading it back.
How to: Delete Files and Directories in Isolated Storage Demonstrates how to delete isolated storage files and directories.
File and Stream I/O Explains how you can perform synchronous and asynchronous file and data stream access.

Reference