Skip to content

JanneMattila/azure-policy-evaluator

Repository files navigation

Azure Policy Evaluator

ci

Caution

This is a personal and very experimental project so don't expect anything from it.

Background

See more detailed background blog post here.

Azure Policy development is too hard. Unfortunately, the typical policy development flow is:

  • You manipulate JSON files
  • You deploy them to Azure
  • You wait a bit
  • You create manually resources for your test scenarios to see if your policy works as expected

Above is not very efficient and it's time consuming and it does not match modern development practices that people are used to.

You might not always even realize that testing just in Azure Portal is not enough. Azure Portal might use APIs in such a way that your blocking policy works, but if you then use e.g., Azure CLI or Azure PowerShell, then your policy might not work as expected.

Example:

Try to deny inbound traffic to port 22 on Network Security Group (NSG).

If you test this policy in Azure Portal by adding inbound rule to allow this traffic, policy correctly blocks you from doing that. So, it works as expected.

If you then try to do the same with Azure PowerShell:

$nsg = Get-AzNetworkSecurityGroup -Name $nsgName -ResourceGroupName $resourceGroupName

$nsg | Add-AzNetworkSecurityRuleConfig `
    -Name $ruleName `
    -Description "Allow SSH" `
    -Access Allow `
    -Protocol "*" `
    -Direction "Inbound" `
    -Priority 100 `
    -SourceAddressPrefix "*" `
    -SourcePortRange "*" `
    -DestinationAddressPrefix "*" `
    -DestinationPortRange $port

$nsg |  Set-AzNetworkSecurityGroup

Your policy might not block this time and your inbound rule is successfully created. This is exactly what you don't want to happen.

Above can happen if another one is sending a full NSG object:

Microsoft.Network/networkSecurityGroups

and the other one is sending just the rule:

Microsoft.Network/networkSecurityGroups/securityRules

To test these small differences, it would take even more time.

What if we could use tool to do local testing of our policies? Something that runs very similar to any unit test framework. You edit the file and voila, you see the results immediately in your console.

What if we had example test cases for you to use? You're developing NSG rules, then here are set of test cases to ease your development.

Experiment

The idea is to create a tool that can evaluate Azure Policy definitions against a given Azure Resource Manager JSON objects.

Example ARM Resource "nsg-allow-ssh-deny.json" for Network Security Group (NSG) allowing port 22 usage
{
  "name": "nsg-app",
  "type": "Microsoft.Network/networkSecurityGroups",
  "location": "northeurope",
  "properties": {
    "securityRules": [
      {
        "type": "Microsoft.Network/networkSecurityGroups/securityRules",
        "properties": {
          "protocol": "*",
          "sourcePortRange": "*",
          "destinationPortRange": "22",
          "sourceAddressPrefix": "*",
          "destinationAddressPrefix": "10.0.0.4",
          "access": "Allow",
          "priority": 4096,
          "direction": "Inbound",
          "sourcePortRanges": [],
          "destinationPortRanges": [],
          "sourceAddressPrefixes": [],
          "destinationAddressPrefixes": []
        }
      }
    ]
  }
}
Example policy "azurepolicy.json" to prevent inbound traffic on defined ports

Policy example has been taken from deny-ports-nsg from Community Policy Repo.

{
  "properties": {
    "mode": "All",
    "policyRule": {
      "if": {
        "anyOf": [
          {
            "allOf": [
              {
                "field": "type",
                "equals": "Microsoft.Network/networkSecurityGroups/securityRules"
              },
              {
                "not": {
                  "field": "Microsoft.Network/networkSecurityGroups/securityRules/sourceAddressPrefix",
                  "notEquals": "*"
                }
              },
              {
                "anyOf": [
                  {
                    "field": "Microsoft.Network/networkSecurityGroups/securityRules/destinationPortRange",
                    "equals": "22"
                  },
                  {
                    "field": "Microsoft.Network/networkSecurityGroups/securityRules/destinationPortRange",
                    "equals": "3389"
                  }
                ]
              }
            ]
          },
          {
            "allOf": [
              {
                "field": "type",
                "equals": "Microsoft.Network/networkSecurityGroups"
              },
              {
                "count": {
                  "field": "Microsoft.Network/networkSecurityGroups/securityRules[*]",
                  "where": {
                    "allOf": [
                      {
                        "field": "Microsoft.Network/networkSecurityGroups/securityRules[*].sourceAddressPrefix",
                        "equals": "*"
                      },
                      {
                        "anyOf": [
                          {
                            "field": "Microsoft.Network/networkSecurityGroups/securityRules[*].destinationPortRange",
                            "equals": "22"
                          },
                          {
                            "field": "Microsoft.Network/networkSecurityGroups/securityRules[*].destinationPortRange",
                            "equals": "3389"
                          }
                        ]
                      }
                    ]
                  }
                },
                "greater": 0
              }
            ]
          }
        ]
      },
      "then": {
        "effect": "Deny"
      }
    }
  }
}

Now we can run our tool to evaluate the policy against the resource:

$ ape -p azurepolicy.json -t nsg-allow-ssh-and-rdp-deny.json

Policy 'azurepolicy' with test 'nsg-allow-ssh-and-rdp-deny' evaluated to 'Deny' which was expected -> PASS

Above verifies, that the policy correctly blocks the resource creation.

Limitations

As this is just an experiment, there are many limitations (list is not even exhaustive):

  • Most of the template functions are not implemented
    • parameters is implemented
  • Most of the data types are not implemented
    • string, int and bool are implemented
  • Most of the policy conditions are not implemented
    • field, count, in, notIn, allOf, anyOf, not, equals, notEquals, contains, greater, greaterOrEquals, less, lessOrEquals, exists, like, notLike are implemented at least partially
  • Aliases are implemented but multiple aliases are not correctly handled
    • [*] array alias is implemented
  • "source": "action" is not implemented (info)

Try it yourself

Download the tool:

  1. Go to Actions
  2. Select latest successful run
  3. Download artifact based on your platform
    • ape-windows for Windows
    • ape-linux for Linux
    • ape-macos for macOS
  4. Extract the artifact
  5. Add the executable to location which is in your PATH environment variable

Ready to use:

$ ape --help

Description:
  Azure Policy Evaluator allows you to evaluate Azure Policy files against test files.
  You can use this to test your policies before deploying them to Azure.

  Tool can be used in 3 different ways:

  1. Evaluate a single policy file against a single test file.
  2. Watch policy changes in the folders and evaluate them against all test files in the sub-folders.
  3. Run all tests from a folder.

  More information can be found here:
  https://github.com/JanneMattila/azure-policy-evaluator

Usage:
  ape [options]

Options:
  -p, --policy <policy>              Policy file to evaluate
  -t, --test <test>                  Test file to use in evaluation
  -w, --watch                        Watch for policy changes
  -f, --watch-folder <watch-folder>  Watch folder path
  -r, --run-tests <run-tests>        Run all tests from path
  --logging <debug|info|trace>       Logging verbosity [default: info]
  --version                          Show version information
  -?, -h, --help                     Show help and usage information

Most common usage is to navigate to the policy folder and then execute it using watch mode:

ape -w

See demo about watch in action:

Azure Policy Evaluator: Quick demo about watch mode

To see debug level logging:

ape -w --logging debug

This enables you to see the policy evaluation steps in detail:

$ ape -p azurepolicy.json -t tests\kv-iprules-with-allow-none.json --logging debug

Started policy evaluation
ParseParameters => Started parsing of parameters
ParseParameters => Parsing parameter 'effect' of type 'string'
ParseParameters => Parsed default value 'Audit'
ParseParameters => Parsed 1 parameters
1 => Started evaluation
1 => 'allOf' started
1 => 2 => Started evaluation
1 => 2 => 'field' started
1 => 2 => Started alias cache population
1 => 2 => Finished alias cache population in 206 ms and 65383 items
1 => 2 => Alias 'type' not found in cache
1 => 2 => Field comparison for 'type' with value 'Microsoft.KeyVault/vaults'
1 => 2 => Property 'type' with value 'Microsoft.KeyVault/vaults' "equals" 'Microsoft.KeyVault/vaults' is 'True'
1 => 2 => 'field' return condition 'True'
1 => 2 => Started evaluation
1 => 2 => 'anyOf' started
1 => 2 => 3 => Started evaluation
1 => 2 => 3 => 'field' started
1 => 2 => 3 => Alias 'Microsoft.KeyVault/vaults/networkAcls.virtualNetworkRules[*].id' found in cache 'properties.networkAcls.virtualNetworkRules[*].id'        
1 => 2 => 3 => Property path 'Microsoft.KeyVault/vaults/networkAcls.virtualNetworkRules[*].id' is alias to 'properties.networkAcls.virtualNetworkRules[*].id'   
1 => 2 => 3 => Property 'properties.networkAcls.virtualNetworkRules[*].id' found
1 => 2 => 3 => Property 'networkAcls.virtualNetworkRules[*].id' found
1 => 2 => 3 => Array evaluation for 'virtualNetworkRules[*].id'
1 => 2 => 3 => 'field' return condition 'False'
1 => 2 => 3 => Started evaluation
1 => 2 => 3 => 'field' started
1 => 2 => 3 => Alias 'Microsoft.KeyVault/vaults/networkAcls.virtualNetworkRules[*].id' found in cache 'properties.networkAcls.virtualNetworkRules[*].id'        
1 => 2 => 3 => Property path 'Microsoft.KeyVault/vaults/networkAcls.virtualNetworkRules[*].id' is alias to 'properties.networkAcls.virtualNetworkRules[*].id'   
1 => 2 => 3 => Property 'properties.networkAcls.virtualNetworkRules[*].id' found
1 => 2 => 3 => Property 'networkAcls.virtualNetworkRules[*].id' found
1 => 2 => 3 => Array evaluation for 'virtualNetworkRules[*].id'
1 => 2 => 3 => 'field' return condition 'False'
1 => 2 => 3 => Started evaluation
1 => 2 => 3 => 'field' started
1 => 2 => 3 => Alias 'Microsoft.KeyVault/vaults/networkAcls.defaultAction' found in cache 'properties.networkAcls.defaultAction'
1 => 2 => 3 => Property path 'Microsoft.KeyVault/vaults/networkAcls.defaultAction' is alias to 'properties.networkAcls.defaultAction'
1 => 2 => 3 => Property 'properties.networkAcls.defaultAction' found
1 => 2 => 3 => Property 'networkAcls.defaultAction' found
1 => 2 => 3 => Field comparison for 'defaultAction' with value 'Deny'
1 => 2 => 3 => Property 'defaultAction' with value 'Deny' "equals" 'Allow' is 'False'
1 => 2 => 3 => 'field' return condition 'False'
1 => 2 => 'anyOf' return condition 'False'
1 => 'allOf' return condition 'False'
Policy evaluation finished with 'False' causing effect 'None'
Policy 'azurepolicy' with test 'kv-iprules-with-allow-none' evaluated to 'None' which was expected -> 'PASS'

To evaluate single policy against single test file:

ape -p azurepolicy.json -t nsg-deny.json

To run all tests from a folder and its sub-folders:

ape -r samples

You can use the above e.g., in GitHub Actions or Azure Pipelines to run your tests. You can see example from this repository CI workflow. You can use the exit code to determine if the tests passed or not and then fail the build if needed.

Food for thought

Wouldn't it be cool to be able to clone e.g., Community Policy Repo and then run all tests from there?

To allow using ape from any folder, you can add it to any folder which is in your PATH environment variable.

In case you want to remove ape from your system, you can just delete the executable. If you have forgotten where you installed it, you can use these commands to find the executable:

Command-prompt:

where ape

PowerShell:

gcm ape | fl

How watch mode works

Azure Policy Evaluator finds all *.json files from the current folder and its sub-folders. If the file name is azurepolicy.json, then it's considered as a policy file. Matching test files are files which have *.json extension and they are in the same folder or in the sub-folders.

Here is the directory structure from this repository samples folder:

├───Compute
│   └───audit-vm-byol-compliance
│       │   azurepolicy.json
│       └───tests
│               linux-vm-none.json
│               windows-vm-audit.json
│               windows-vm-with-license-none.json
├───Key Vault
│   └───audit-if-key-vault-has-no-virtual-network-rules
│       │   azurepolicy.json
│       └───tests
│               kv-iprules-with-allow-none.json
│               kv-iprules-with-deny-none.json
│               kv-no-rules-allow-audit.json
│               kv-virtualnetworkrules-with-allow-audit.json
│               kv-virtualnetworkrules-with-deny-none.json
└───Network
    ├───deny-ports-nsg
    │   │   azurepolicy.json
    │   └───tests
    │           nsg-allow-ssh-and-rdp-deny.json
    │           securityrule-allows-ssh-deny.json
    └───enforce-load-balancer-standard-sku
        │   azurepolicy.json
        └───tests
                basic-loadbalancer-audit.json
                standard-loadbalancer-none.json

In tests folder there are test files which are used to evaluate the policy. Test file name is used to describe the test case expected result. E.g., securityrule-allows-ssh-deny.json means that the test case expects the policy to deny the resource.

You can start the watch mode from the root of this repository:

ape -w -f samples

If you now edit any of the policy files or test files, then the tool will automatically run the evaluation again.

If you edit a test file, then the tool will run only that test case.

If you edit a policy file, then the tool will run all test cases which are related to that policy file.

How to create test files

The test files are just ARM resource JSON files.

The easiest way to create test files is to copy existing resource from Azure Portal and then modify it to match your test case. You can use JSON View in the Azure Portal to copy the JSON from any resource.

Remember to use the latest available API Version to get all the relevant fields. Sometimes it defaults to an older API Version which might not contain all the fields you need. You can remove any extra fields, identifiers and others which are not needed for the test case.

Last part of the test file name is used to describe the expected policy evaluation result. E.g., keyvault-no-rules-audit.json means that the test case expects the policy to audit that specific resource. Name seperator is - and it's used to split the test case name from the expected result.

Example policy and test cases

If you need to create policy for denying inbound traffic to port 22 (SSH) or 3389 (RDP), then you're looking to implement Network Security Group (NSG) policy.

You can find example policy from deny-ports-nsg

To test this policy, you can use these test cases:

Feedback

Use GitHub Discussions to give feedback or provide your comments and ideas. It would be great to hear your thoughts about this tool and that do you see value in it.

I hope this tool brings some thoughts and ideas to you. Would you use this to develop more policies since it's so easy to test them?

The main question is, do you see value in this tool?

Links

Azure Policy Samples

Community Policy Repo

About

Azure Policy development should be just like normal development with unit testing capabilities. This project tries to enable that.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published