How WDAC works with PowerShell

發行項
09/23/2024

This article explains how Windows Defender Application Control (WDAC) secures PowerShell and the restrictions it imposes. The secure behavior of PowerShell varies based on the version of Windows and PowerShell you're using.

How PowerShell detects a system lockdown policy

PowerShell detects both AppLocker and Windows Defender Application Control (WDAC) system wide polices. AppLocker is deprecated. WDAC is the preferred application control system for Windows.

Legacy WDAC policy enforcement detection

PowerShell uses the legacy WDAC WldpGetLockdownPolicy API to discover two things:

System wide policy enforcement: None, Audit, Enforce
Individual file policy: None, Audit (allowed by policy), Enforce (not allowed by policy)

All versions of PowerShell (v5.1 - v7.x) support this WDAC policy detection.

Latest WDAC policy enforcement detection

WDAC introduced new APIs in recent versions of Windows. Beginning with version 7.3, PowerShell uses the new WldpCanExecuteFile API to decide how a file should be handled. Windows PowerShell 5.1 doesn't support this new API. The new API takes precedence over the legacy API for individual files. However, PowerShell continues to use the legacy API to get the system wide policy configuration. If the new API isn't available, PowerShell falls back to the old API behavior.

The new API provides the following information for each file:

WLDP_CAN_EXECUTE_ALLOWED
WLDP_CAN_EXECUTE_BLOCKED
WLDP_CAN_EXECUTE_REQUIRE_SANDBOX

PowerShell behavior under lockdown policy

PowerShell can run in both interactive and non-interactive modes.

In interactive mode, PowerShell is a command-line application that takes users command-line input as commands or scripts to run. Results are displayed back to the user.
In non-interactive mode, PowerShell loads modules and runs script files without user input. Result data streams are either ignored or redirected to a file.

Interactive mode running under policy enforcement

PowerShell runs commands in ConstrainedLanguage mode. This mode prevents interactive users from running certain commands or executing arbitrary code. For more information about the restrictions in this mode, see the PowerShell restrictions under lockdown policy section of this article.

Noninteractive mode running under policy enforcement

When PowerShell runs a script or loads a module, it uses the WDAC API to get the policy enforcement for the file.

PowerShell version 7.3 or higher uses the WldpCanExecuteFile API if available. This API returns one of the following results:

WLDP_CAN_EXECUTE_ALLOWED: The file is approved by policy and is used in FullLanguage mode with a few restrictions.
WLDP_CAN_EXECUTE_BLOCKED: The file isn't approved by policy. PowerShell throws an error when the file is run or loaded.
WLDP_CAN_EXECUTE_REQUIRE_SANDBOX: The file isn't approved by the policy but it can still be run or loaded in ConstrainedLanguage mode.

In Windows PowerShell 5.1 or if WldpCanExecuteFile API isn't available, PowerShell's per file behavior is:

None: The file is run loaded in FullLanguage mode with a few restrictions.
Audit: The file is run or loaded in FullLanguage mode with no restrictions. In PowerShell 7.4 or higher, the policy logs restriction information to the Windows event logs.
Enforce: The file is run or loaded in ConstrainedLanguage mode.

PowerShell restrictions under lockdown policy

When PowerShell detects the system is under a WDAC lockdown policy, it applies restrictions even if the script is trusted and running in FullLanguage mode. These restrictions prevent known behaviors of PowerShell that could result in arbitrary code execution on a locked-down system. The lockdown policy enforces the following restrictions:

Module dot-sourcing with wildcard function export restriction

Any module that uses script dot-sourcing and exports functions using wildcard names results in an error. Blocking wildcard exports prevents script injection from a malicious user who can plant an untrusted script that gets dot-sourced into a trusted module. The malicious script could then gain access to the trusted module's private functions.

Security recommendation: Never use script dot-sourcing in a module and always export module functions with explicit names (no wildcard characters).
Nested module with wildcard function export restriction

If a parent module exports functions using function name wildcard characters, PowerShell removes any function name in a nested module from the function export list. Blocking wildcard exports from nested modules prevents accidental exporting of dangerous nested functions through wildcard name matching.

Security recommendation: Always export module functions with explicit names (no wildcard characters).
Interactive shell parameter type conversion

When the system is locked down, interactive PowerShell sessions run in ConstrainedLanguage mode to prevent arbitrary code execution. Trusted modules loaded into the session run in FullLanguage mode. If a trusted module cmdlet uses complex types for its parameters, type conversion during parameter binding can fail if the conversion isn't allowed across trust boundaries. The failure occurs when PowerShell tries to convert a value by running a type constructor. Type constructors aren't allowed to run in ConstrainedLanguage mode.

In this example, parameter binding type conversion is normally allowed, but fails when run in ConstrainedLanguage mode. The ConnectionPort type constructor isn't allowed:
```
PS> Create-ConnectionOnPort -Connection 22
Create-ConnectionOnPort: Cannot bind parameter 'Connection'. Cannot convert the "22"
value of type "System.Int32" to type "ConnectionPort".
```
Enter-PSHostProcess cmdlet disallowed

The Enter-PSHostProcess cmdlet is disabled and throws an error if used. This command is used for attach-and-debug sessions. It allows you to connect to any other PowerShell session on the local machine. The cmdlet is disabled to prevent information disclosure and arbitrary code execution.

PowerShell restrictions under constrained language mode

Script or function that isn't approved by the WDAC policy is untrusted. When you run an untrusted command, PowerShell either blocks the command from running (new behavior) or runs the command in ConstrainedLanguage mode. The following restrictions apply to ConstrainedLanguage mode:

Add-Type cmdlet disallowed

Blocking Add-Type prevents the execution of arbitrary .NET code.
Import-LocalizedData cmdlet restricted

Blocking the SupportedCommand parameter of Import-LocalizedData prevents the execution of arbitrary code.
Invoke-Expression cmdlet restricted

All script blocks passed to the Invoke-Expression cmdlet are run in ConstrainedLanguage mode to prevent arbitrary code execution.
New-Object cmdlet restricted

The New-Object cmdlet is restricted to use only allowed .NET and COM types, to prevent access to untrusted types.
ForEach-Object cmdlet restriction

Type method invocation for variables passed to the ForeEach-Object is disallowed for any .NET type not in the approved list. In general, ConstrainedLanguage mode disallows any object method invocation except for approved .NET types to prevent access to untrusted .NET types.
Export-ModuleMember cmdlet restriction

Using Export-ModuleMember cmdlet to export functions in a nested module script file where the child module isn't trusted and the parent module is trusted, results in an error. Blocking this scenario prevents a malicious untrusted module from exporting dangerous functions from a trusted module.
New-Module cmdlet restriction

When you run New-Module in a trusted script, any script block provided by the ScriptBlock parameter is marked to run in ConstrainedLanguage mode to prevent the injection of arbitrary code into a trusted execution context.
Configuration keyword not allowed

The Configuration language keyword isn't allowed in ConstrainedLanguage mode to prevent code injection attacks.
class keyword not allowed

The class language keyword isn't allowed in ConstrainedLanguage mode to prevent the injection of arbitrary code.
Script Block processing scope restrictions

Child script blocks aren't allowed to run in parent script block scopes if the script blocks have different trust levels. For example, you create a child relationship when you dot-source one script into another. Blocking this scenario prevents an untrusted script from getting access to dangerous functions in the trusted script scope.
Prevent command discovery of untrusted script functions

PowerShell command discovery doesn't return functions from an untrusted source, such as an untrusted script or module, to a trusted function. Blocking discovery of untrusted commands prevents code injection through command planting.
Hashtable to object conversion not allowed

ConstrainedLanguage mode blocks hashtable to object conversions in the Data sections of PowerShell data (.psd1) files to prevent potential code injection attacks.
Automatic type conversion restricted

ConstrainedLanguage mode blocks automatic type conversion except for a small set of approved safe types to prevent potential code injection attacks.
Implicit module function export restriction

If a module doesn't explicitly export functions, PowerShell implicitly exports all defined module functions automatically as a convenience feature. In ConstrainedLanguage mode, implicit exports no longer occur when a module is loaded across trust boundaries. Blocking implicit exports prevents unintended exposure of dangerous module functions not meant for public use.
Script files can't be imported as modules

PowerShell allows you to import script files (.ps1) as a module. All defined functions become publicly accessible. ConstrainedLanguage mode blocks importation of script file to prevent unintended exposure of dangerous script functions.
Setting variables AllScope restriction

ConstrainedLanguage mode disables the ability to set AllScope on variables. Limiting the scope of variables prevents the variables from interfering with the session state of trusted commands.
Type method invocation not allowed

ConstrainedLanguage mode doesn't allow method invocation on unapproved types. Blocking methods on unapproved types prevents invocation of .NET type methods that might be dangerous or allow code injection.
Type property setters not allowed

ConstrainedLanguage mode restricts invocation of property setters on unapproved types. Blocking property setters on unapproved types prevents code injection attacks.
Type creation not allowed

ConstrainedLanguage mode blocks type creation on unapproved types to block untrusted constructors that could allow code injection.
Module scope operator not allowed

ConstrainedLanguage mode doesn't allow the use of the module scope operator. For example: & (Get-Module MyModule) MyFunction. Blocking the module scope operator prevents access to module private functions and variables.

共用方式為