| Package | Affected versions | Patched versions |
|---|---|---|
| github.com/open-policy-agent/opa | < 1.4.0 | 1.4.0 |
| github.com/open-policy-agent/opa/server | < 1.4.0 | 1.4.0 |
| github.com/open-policy-agent/opa/v1/server | < 1.4.0 | 1.4.0 |
When run as a server, OPA exposes an HTTP Data API for reading and writing documents. Requesting a virtual document through the Data API entails policy evaluation, where a Rego query containing a single data document reference is constructed from the requested path. This query is then used for policy evaluation.
A HTTP request path can be crafted in a way that injects Rego code into the constructed query. The evaluation result cannot be made to return any other data than what is generated by the requested path, but this path can be misdirected, and the injected Rego code can be crafted to make the query succeed or fail; opening up for oracle attacks or, given the right circumstances, erroneous policy decision results. Furthermore, the injected code can be crafted to be computationally expensive, resulting in a Denial Of Service (DoS) attack.
Users are only impacted if all of the following apply:
input.path attribute when deciding if the request should be allowed.or, if all of the following apply:
Note: With no Authorization Policy configured for restricting API access (the default configuration), the RESTful Data API provides access for managing Rego policies; and the RESTful Query API facilitates advanced queries. Full access to these APIs provides both simpler, and broader access than what the security issue describes here can facilitate. As such, OPA servers exposed to a network are not considered affected by the attack described here if they are knowingly not restricting access through an Authorization Policy.
Fixed in OPA v1.4.0.
Unless necessary for production reasons, network access to OPA’s RESTful APIs should be limited to localhost and/or trusted networks.
Since OPA v1.0, unless otherwise configured, the server listener defaults to .
Running one of these products? Scan your site to see whether it exposes a vulnerable version.
Scan my site freelocalhostA configured authentication scheme is a requirement when OPA is exposed in an untrusted environment. While requiring authentication alone doesn’t mitigate this attack, it effectively reduces the scope from untrusted clients to trusted clients.
OPA can be configured to use an Authorization Policy to validate all incoming requests. By authoring the Authorization Policy to only accept paths corresponding to expected Rego package references, this attack can be fully mitigated.
The HTTP path in a Data API request is of the format /v1/data/{path:.+} (/v0/data/{path:.+}, for the v0 Data API), where data/{path:.+} directly corresponds to a reference to a virtual document, and a prefix of {path:.+} corresponds to a Rego package declaration.
E.g. the HTTP path v1/data/do/re/mi corresponds to the data reference data.do.re.mi, where do.re is the package and mi is the rule in the following Rego module:
package do.re
mi if {
...
}
Unless otherwise configured, OPA will use the rule at data.system.authz.allow as Authorization Policy. Authorization is enabled by starting OPA with the --authorization=basic flag, and the Authorization policy must be made available to the OPA runtime either through a bundle (via the --bundle flag or through discovery) or as an individual module via the command-line.
A trivial Authorization Policy example:
package system.authz
allowed_paths := [
["v1", "data", "policy1", "allow"],
["v1", "data", "policy2", "allow"],
...
]
allow if {
input.path in allowed_paths
}
Note: configuring an Authorization Policy in OPA isn't the only way to protect against malicious request paths. Path validation and sanitisation can also be performed by connecting clients and 3rd party intermediaries, such as API gateways, reverse proxies, etc.