Skip to content

OPA Compiler: Bypass of WithUnsafeBuiltins using "with" keyword to mock functions

High severity GitHub Reviewed Published Sep 7, 2022 in open-policy-agent/opa • Updated Jan 23, 2024

Package

gomod github.com/open-policy-agent/opa (Go)

Affected versions

>= 0.40.0, < 0.43.1

Patched versions

0.43.1

Description

Impact

The Rego compiler provides a (deprecated) WithUnsafeBuiltins function, which allows users to provide a set of built-in functions that should be deemed unsafe — and as such rejected — by the compiler if encountered in the policy compilation stage. A bypass of this protection has been found, where the use of the with keyword to mock such a built-in function (a feature introduced in OPA v0.40.0), isn’t taken into account by WithUnsafeBuiltins.

The same method is exposed via rego.UnsafeBuiltins in the github.com/open-policy-agent/opa/rego package.

When provided e.g. the http.send built-in function to WithUnsafeBuiltins, the following policy would still compile, and call the http.send function with the arguments provided to the is_object function when evaluated:

package policy

foo := is_object({
    "method": "get", 
    "url": "https://www.openpolicyagent.org"
})

allow := r {
    r := foo with is_object as http.send
}

Both built-in functions and user provided (i.e. custom) functions are mockable using this construct.

In addition to http.send, the opa.runtime built-in function is commonly considered unsafe in integrations where policy provided by untrusted parties is evaluated, as it risks exposing configuration, or environment variables, potentially carrying sensitive information.

Affected Users

All of these conditions have to be met to create an adverse effect:

  • Use the Go API for policy evaluation (not the OPA server, or the Go SDK)
  • Make use of the WithUnsafeBuiltins method in order to deny certain built-in functions, like e.g. http.send, from being used in policy evaluation.
  • Allow policy evaluation of policies provided by untrusted parties.
  • The policies evaluated include the with keyword to rewrite/mock a built-in, or custom, function to that of another built-in function, such as http.send.

Additionally, the OPA Query API is affected:

  • If the OPA Query API is exposed to the public, and it is relied on http.send to be unavailable in that context. Exposing the OPA API to the public without proper authentication and authorization in place is generally advised against.

Patches

v0.43.1, v0.44.0

Workarounds

The WithUnsafeBuiltins function has been considered deprecated since the introduction of the capabilities feature in OPA v0.23.0 . While the function was commented as deprecated, the format of the comment was however not following the convention for deprecated functions, and might not have been picked up by tooling like editors. This has now been fixed. Users are still encouraged to use the capabilities feature over the deprecated WithUnsafeBuiltins function.

If you cannot upgrade, consider using capabilities instead:

Code like this using the github.com/open-policy-agent/opa/ast package:

// VULNERABLE with OPA <= 0.43.0
unsafeBuiltins := map[string]struct{}{
	ast.HTTPSend.Name: struct{}{},
}
compiler := ast.NewCompiler().WithUnsafeBuiltins(unsafeBuiltins)

needs to be changed to this:

caps := ast.CapabilitiesForThisVersion()
var j int
for i, bi := range caps.Builtins {
	if bi.Name == ast.HTTPSend.Name {
		j = i
		break
	}
}
caps.Builtins[j] = caps.Builtins[len(caps.Builtins)-1] // put last element into position j
caps.Builtins = caps.Builtins[:len(caps.Builtins)-1]   // truncate slice

compiler := ast.NewCompiler().WithCapabilities(caps)

When using the github.com/open-policy-agent/opa/rego package:

// VULNERABLE with OPA <= 0.43.0
r := rego.New(
	// other options omitted
	rego.UnsafeBuiltins(map[string]struct{}{ast.HTTPSend.Name: struct{}{}}),
)

needs to be changed to

r := rego.New(
	// other options omitted
	rego.Capabilities(caps),
)

with caps defined above.

Note that in the process, some error messages will change: http.send in this example will no longer be "unsafe" and thus forbidden, but it will simply become an "unknown function".

References

For more information

If you have any questions or comments about this advisory:

References

@srenatus srenatus published to open-policy-agent/opa Sep 7, 2022
Published by the National Vulnerability Database Sep 8, 2022
Published to the GitHub Advisory Database Sep 16, 2022
Reviewed Sep 16, 2022
Last updated Jan 23, 2024

Severity

High

CVSS overall score

This score calculates overall vulnerability severity from 0 to 10 and is based on the Common Vulnerability Scoring System (CVSS).
/ 10

CVSS v3 base metrics

Attack vector
Network
Attack complexity
High
Privileges required
None
User interaction
None
Scope
Unchanged
Confidentiality
High
Integrity
High
Availability
None

CVSS v3 base metrics

Attack vector: More severe the more the remote (logically and physically) an attacker can be in order to exploit the vulnerability.
Attack complexity: More severe for the least complex attacks.
Privileges required: More severe if no privileges are required.
User interaction: More severe when no user interaction is required.
Scope: More severe when a scope change occurs, e.g. one vulnerable component impacts resources in components beyond its security scope.
Confidentiality: More severe when loss of data confidentiality is highest, measuring the level of data access available to an unauthorized user.
Integrity: More severe when loss of data integrity is the highest, measuring the consequence of data modification possible by an unauthorized user.
Availability: More severe when the loss of impacted component availability is highest.
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:N

EPSS score

0.301%
(70th percentile)

Weaknesses

CVE ID

CVE-2022-36085

GHSA ID

GHSA-f524-rf33-2jjr

Source code

Credits

Loading Checking history
See something to contribute? Suggest improvements for this vulnerability.