Troubleshooting Google Cloud Armor issues

Use these instructions to troubleshoot issues with your Google Cloud Armor security policies.

General issues

Traffic is allowed despite a deny rule configured in the Google Cloud Armor security policy

To fix this, follow these steps:

Make sure that the Google Cloud Armor security policy is attached to a target backend service. For example, the following command describes all data associated with the backend service BACKEND. The results returned should include the name of the Google Cloud Armor security policy associated with this backend service.
```
gcloud compute backend-services describe BACKEND
```

Review the HTTP(S) logs to determine which policy and rule were matched for your traffic along with the associated action. To view the logs, use Cloud Logging.

The following is a sample log of an allowed request with the interesting fields highlighted. Check for the following fields and make sure that they match the rule that you configured to deny the traffic:

configuredAction should match the action configured in the rule.
name should match the name of the Google Cloud Armor security policy attached to this backend service.
outcome should match configuredAction.
priority should match the priority number of the rule.

  httpRequest:
   remoteIp: 104.133.0.95
   requestMethod: GET
   requestSize: '801'
   requestUrl: http://74.125.67.38/
   responseSize: '246'
   serverIp: 10.132.0.4
   status: 200
   userAgent: curl/7.35.0
     insertId: ajvis5ev4i60
     internalId:
       projectNumber: '895280006100'
     jsonPayload:
       '@type': type.googleapis.com/google.cloud.loadbalancing.type.LoadBalancerLogEntry
       enforcedSecurityPolicy:
         configuredAction: ACCEPT
         name: mydev-policy-log-test1
         outcome: ACCEPT
         priority: 2147483647
       statusDetails: response_sent_by_backend
     logName: projects/mydev-staging/logs/requests
     resource:
       labels:
         backend_service_name: BACKEND_SERVICE_NAME
         forwarding_rule_name: FORWARDING_RULE_NAME
         project_id: PROJECT_ID
         target_proxy_name: TARGET_HTTP_PROXY_NAME
         url_map_name: URL_MAP_NAME
         zone: global
       type: http_load_balancer
     severity: INFO
     timestamp: '2017-04-18T18:57:05.845960288Z'

Review the hierarchy of rules to ensure that the correct rule is matched. It is possible that a higher priority rule with an allow action is matching your traffic. Use the describe command on the security-policies in the gcloud command-line tool to see the contents of the Google Cloud Armor security policy.

For example, the following example shows how a higher priority allow rule (at priority 100) matches traffic coming from the 1.2.3.4 IP address, preventing the lower priority deny rule (at priority 200) from triggering and blocking the traffic.

gcloud compute security-policies describe POLICY_NAME

Output:

  creationTimestamp: '2017-04-18T14:47:58.045-07:00
  description: ''
  fingerprint: Yu5spBjdoC0=
  id: '2560355463394441057'
  kind: compute#securityPolicy
  name: POLICY_NAME
  rules:
  -action: allow
   description: allow high priority rule
   kind: compute#securityPolicyRule
   match:
     srcIpRanges:
     -'1.2.3.4/32'
   preview: false
   priority: 100
  -action: deny
   description: deny lower priority rule
   kind: compute#securityPolicyRule
   match:
     srcIpRanges:
     -'1.2.3.0/24
   preview: false
   priority: 200
  -action: deny
   description: default rule
   kind: compute#securityPolicyRule
   match:
     srcIpRanges:
     -'*'
   preview: false
   priority: 2147483647
   selfLink: http://www.googleapis.com/compute/v1/projects/bigclustertestdev0-devconsole/global/securityPolicies/sp

Preconfigured rule returns false positives

XSS and SQLi detection are based on static signature matching on HTTP request headers and other L7 parameters. These regular expression patterns are prone to false positives. You can use the preconfigured rule for XSS and SQLi detection in preview mode and then check the log for any false positives.

If you find a false positive, you can compare the traffic content with the ModSecurity CRS rules. If the rule is invalid or not relevant, disable it by using the evaluatePreconfiguredExpr expression, and specify the rule's ID in the exclude ID list argument.

After reviewing the logs and removing all false positives, disable the preview mode.

To add a preconfigured rule in preview mode:

Create a security policy with the preconfigured expression set in preview mode:

gcloud compute security-policies rules create 1000
   --security-policy POLICY_NAME
   --expression "evaluatePreconfiguredExpr('xss-stable')"
   --action deny-403
   --preview

Review the HTTP(S) logs for HTTP request fields such as url and cookie. For example, the requestUrl compares positively to the ModSecurity CRS rule ID 941180:

httpRequest:
  remoteIp: 104.133.0.95
  requestMethod: GET
  requestSize: '801'
  requestUrl: http://74.125.67.38/foo?document.cookie=1010"
  responseSize: '246'
  serverIp: 10.132.0.4
  status: 200
  userAgent: curl/7.35.0
insertId: ajvis5ev4i60
internalId:
  projectNumber: '895280006100'
jsonPayload:
  '@type': type.googleapis.com/google.cloud.loadbalancing.type.LoadBalancerLogEntry
  enforcedSecurityPolicy:
    configuredAction: ACCEPT
    name: POLICY_NAME
    outcome: ACCEPT
    priority: 2147483647
    preconfiguredExprIds: [ 'owasp-crs-v030001-id941180-xss' ]
  statusDetails: response_sent_by_backend
logName: projects/mydev-staging/logs/requests
resource:
  labels:
    backend_service_name: BACKEND_SERVICE
    forwarding_rule_name: mydev-forwarding-rule
    project_id: mydev-staging
    target_proxy_name: mydev-target-http-proxy
    url_map_name: mydev-url-map
    zone: global
  type: http_load_balancer
severity: INFO
timestamp: '2017-04-18T18:57:05.845960288Z'

Exclude the ModSecurity CRS rule ID 941180 by updating the rule in the Google Cloud Armor security policy:

gcloud compute security-policies rules update 1000 \
    --security-policy POLICY_NAME \
    --expression "evaluatePreconfiguredExpr('xss-stable', ['owasp-crs-v030001-id941180-xss'])" \
    --action deny-403 \
    --preview

Review the logs again and then disable preview mode to implement the rule.

Clients with denied signatures are not blocked or denied

If you are using Google Cloud Armor with Cloud CDN, security policies are enforced only for requests for dynamic content, cache misses, or other requests that are destined for the CDN origin server. Cache hits are served even if the downstream Google Cloud Armor security policy would prevent that request from reaching the CDN origin server.

Issues with named IP address lists

This section provides information for resolving problems with named IP address lists.

IP addresses within a named IP address list

The IP addresses in the lists always match the IP addresses in the provider websites listed in the Google Cloud Armor named IP address lists guide. If you have questions about the lists, contact the Cloud Support team.

IP addresses within a named IP address list are stale in Google Cloud Armor

Google Cloud Armor synchronizes its lists with IP address list providers daily. It is possible to have stale data that lags a few hours or a day behind the data at a provider. However, if you believe that the stale data lags more than expected, for example, more than a week, contact the Cloud Support team.

Difficulty creating a security policy that references a named IP address list

You might try to create a security policy that references a named IP address list, using a command such as this:

gcloud compute security-policies rules create 750 \
    --security-policy my \
    --expression "evaluatePreconfiguredExpr('sourceiplist-abc')" \
    --action "allow"

If the command fails, the error that you see looks something like this:

ERROR: (gcloud.compute.security-policies.rules.create) Could not fetch resource:
 - Invalid value for field 'resource.match.expr': '{  "expression": "evaluatePreconfiguredExpr(\u0027sourceiplist-abc\u0027)"}'. Error parsing Cloud Armor rule matcher expression: sourceiplist-abc is not a valid preconfigured expression set.

Make sure that the particular provider is supported, and that the name of the IP address list is given correctly. You can check this by using the following gcloud command to list the current preconfigured expression sets:

gcloud compute security-policies list-preconfigured-expression-sets

Traffic is blocked despite a preconfigured rule for a named IP address allowlist

You might find that traffic is blocked from an IP address that is on a named IP address list:

Make sure that traffic is coming from an IP address that is on a named IP address allowlist.
Check whether there are other security rules with a higher priority that can block the traffic. If you still see the issue, contact the Cloud Support team.
Make sure that the security policy is attached to the correct backend service:
```
gcloud compute backend-services describe BACKEND_SERVICE
```

Check the rules that are in the security policy. For example:

 gcloud compute security-policies describe POLICY_NAME

The command returns information similar to the following:

---
…
name: POLICY_NAME
rules:
-action: allow
 description: allow fastly ip addresses
 kind: compute#securityPolicyRule
 match:
    expr:
      expression: evaluatePreconfiguredExpr('sourceiplist-fastly')
 preview: false
 priority: 100
-action: deny(403)
 description: Default rule, higher priority overrides it
 kind: compute#securityPolicyRule
 match:
    config:
      srcIpRanges:
      -'*'
    versionedExpr: SRC_IPS_V1
 preview: false
 priority: 2147483647
-action: deny(404)
 description: deny xyz referer
 kind: compute#securityPolicyRule
 match:
    expr:
      expression: request.headers['Referer'].contains('xyz')
 preview: false
 priority: 50
…

The preceding security policy contains three rules: a default deny rule, an allow rule for Fastly's IP addresses, and a deny rule for a referrer that contains xyz. Their respective priorities are also listed.

Review the logs to determine which rule is matched to your traffic and the associated action. For information about logging, see Viewing logs in the Legacy Logs Viewer for the Cloud Logging API.

The following is an example of a log:
```
 httpRequest: {
    referer: "http://www.xyz.com/"
    remoteIp: "23.230.32.10"
    requestMethod: "HEAD"
    requestSize: "79"
    requestUrl: "http://www.abc.com/"
    responseSize: "258"
    status: 404
    userAgent: "Mozilla/5.0"
  }
  …
  jsonPayload: {
    @type: "type.googleapis.com/google.cloud.loadbalancing.type.LoadBalancerLogEntry"
    enforcedSecurityPolicy: {
      configuredAction: "DENY"
      name: "POLICY_NAME"
      outcome: "DENY"
      priority: 50
    }
    statusDetails: "denied_by_security_policy"
  }
  …
```
From the preceding log, the request is coming from 23.230.32.10, which is covered by Fastly's public IP address list. However, the request is matched with a deny rule of a higher priority of 50. Comparing this with what's in the security policy, the rule corresponds to the deny rule for a referer that contains xyz. Because the request has a referer of http://www.xyz.com/, the security rule enforcement is working correctly.

Issues with Security Command Center findings

The Google Cloud Armor card does not appear in the Security Command Center

Enable Google Cloud Armor findings in the Security Command Center interface.

Findings from Google Cloud Armor do not appear in the Security Command Center

If findings from Google Cloud Armor do not appear in the Security Command Center, traffic to the backend services might not meet the criteria for raising a finding.

For questions about traffic volume to your backends, check request statistics in the Cloud Monitoring dashboards under the Network Security Policies.

Findings are too noisy

For assistance with this issue, contact the Cloud Support team.

What's next

Get support