OneFirewall vs Google Cloud Armor: one deny list, every backend.
Cloud Armor protects Google Cloud load balancers with WAF rules and DDoS defence. OneFirewall keeps a Cloud Armor address group filled with Alliance-confirmed attacker IPs, so every security policy that references it blocks them, without re-running Terraform.
WAF & cloud edge · Last reviewed: 20 November 2026
Cloud Armor and OneFirewall are complementary, focused on the WAF layer. Cloud Armor evaluates requests at Google's edge with preconfigured WAF rules, rate limiting, Adaptive Protection and (on Enterprise) Google's own threat intelligence lists. OneFirewall adds an independent attacker list from 290+ organisations, maintained inside an address group that all your policies share.
Security policies on Google's external load balancers
Google Cloud Armor attaches security policies to backend services behind Google's external Application Load Balancers. Each policy is an ordered list of rules: preconfigured WAF rules based on the OWASP ModSecurity Core Rule Set, custom rules in Common Expression Language, rate limiting, and bot management with reCAPTCHA. Adaptive Protection uses machine learning to spot layer-7 DDoS.
Cloud Armor Enterprise adds threat intelligence lists from Google and address groups: centrally managed IP lists that many policies can reference with evaluateAddressGroup(), updated through the API or gcloud.
- Enforced at Google's edge, before traffic reaches your backends
- WAF rules inspect request content; IP lists answer who is asking
- Address groups require the Enterprise tier and a fixed capacity
- Updates via API or gcloud apply to every policy that references the group
An address group the WCF Agent keeps current
The integration uses the documented Cloud Armor Enterprise address group pattern. You create the group and the deny rule once; the WCF Agent maintains the contents.
Create the address group
Create a google_network_security_address_group with purpose CLOUD_ARMOR (google-beta provider). Choose the capacity carefully: it can't be changed later. CLOUD_ARMOR-only groups can hold 10,000+ IPs.
Add the deny rule
In each security policy, add a high-priority rule with evaluateAddressGroup(..., origin.ip) and action deny(403). Behind a CDN, set user_ip_request_headers and match origin.user_ip instead.
Grant the agent access
Give the WCF Agent's service account compute.securityAdmin and compute.networkAdmin, or a narrower custom role.
Automatic sync
The agent adds IPs above your Crime Score threshold with add-items and removes decayed ones with remove-items. Every policy that references the group picks up the change.
What the configuration looks like
# Cloud Armor Enterprise: deny rule referencing the OneFirewall address group
expression = "evaluateAddressGroup('projects/<p>/locations/global/addressGroups/denylist-addresses', origin.ip)"
action = "deny(403)" # use origin.user_ip behind a CDN (X-Forwarded-For)
# The WCF Agent keeps the group current without re-running Terraform:
gcloud network-security address-groups add-items denylist-addresses \
--location global --items "203.0.113.47/32,198.51.100.23/32"
gcloud network-security address-groups remove-items denylist-addresses \
--location global --items "198.51.100.23/32" # score decayed
If a project is downgraded from Enterprise, policies that reference address groups become read-only until those rules are removed. Plan the tier before you roll out.
Cloud Armor and OneFirewall, capability by capability
| Capability | Cloud Armor | OneFirewall |
|---|---|---|
| Question it answers | Does this request match a WAF rule, rate limit or DDoS pattern? | Is this client a known attacker right now? |
| Intelligence | Google threat intelligence lists (Enterprise) | Alliance list from 290+ organisations, partners and a honeynet, independent of Google |
| Where it lives | Rules inside each security policy | One address group referenced by all policies |
| Updating | Terraform, API or console changes | Automatic add and remove by the WCF Agent; no Terraform run |
| Scope | Google Cloud load balancers | The same list feeds firewalls, Cloudflare, AWS WAF and SIEMs |
| Behind a CDN | origin.user_ip with trusted headers | Fully supported in the same deny rule |
What OneFirewall adds to Cloud Armor
A source Google doesn't own
Alliance data comes from members' firewalls, VPNs and servers across many providers, a different view from Google's own lists.
Infrastructure as code without the churn
The address group is declared once in Terraform; its contents change every sync without noisy plans or pipeline runs.
Consistent blocking on hybrid estates
The attacker blocked at Cloud Armor is blocked on your on-prem firewall and your other clouds by the same feed.
- Preconfigured WAF rules for SQL injection, XSS and other OWASP risks
- Adaptive Protection for layer-7 DDoS
- Rate limiting and reCAPTCHA bot management
- Google's own threat intelligence lists on the Enterprise tier
Full Terraform examples are in the Cloud Armor integration guide.
OneFirewall and Cloud Armor: FAQ
Do I need Cloud Armor Enterprise?
Yes. Address groups with purpose CLOUD_ARMOR are only available to projects enrolled in Cloud Armor Enterprise.
How many IPs can the address group hold?
CLOUD_ARMOR-only groups can hold 10,000 or more IPs with a quota increase; groups that also use DEFAULT are limited to 1,000. Capacity can't be changed after creation.
Do I have to run Terraform for every update?
No. Terraform creates the group and the rule once. The WCF Agent updates the contents through gcloud or the API.
What if my traffic comes through a CDN?
Configure user_ip_request_headers (for example X-Forwarded-For) and match origin.user_ip in the deny rule.
