Monitoring and tuning - Amazon WAF, Amazon Firewall Manager, and Amazon Shield Advanced
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Monitoring and tuning

This section describes how to monitor and tune your Amazon WAF protections.


To follow the guidance in this section, you need to understand generally how to create and manage Amazon WAF protections like web ACLs, rules, and rule groups. That information is covered in earlier sections of this guide.

Monitor web traffic and rule matches to verify the behavior of the web ACL. If you find problems, adjust your rules to correct and then monitor to verify the adjustments.

Repeat the following procedure until the web ACL is managing your web traffic as you need it to.

To monitor and tune
  1. Monitor traffic and rule matches

    Make sure that traffic is flowing and that your test rules are finding matching requests.

    Look for the following information for the protections that you're testing:

    • Logs – Access information about the rules that match a web request:

      • Your rules - Rules in the web ACL that have Count action are listed under nonTerminatingMatchingRules. Rules with Allow or Block are listed as the terminatingRule. Rules with CAPTCHA or Challenge can be either terminating or non-terminating, and so are listed under one of the two categories, according to the result of the rule match.

      • Rule groups - Rule groups are identified in the ruleGroupId field, with their rule matches categorized the same as for standalone rules.

      • Labels - Labels that rules have applied to the request are listed in the Labels field.

      For more information, see Log fields.

    • Amazon CloudWatch metrics – You can access the following metrics for your web ACL request evaluation.

      • Your rules – Metrics are grouped by the rule action. For example, when you test a rule in Count mode, its matches are listed as Count metrics for the web ACL.

      • Your rule groups – The metrics for your rule groups are listed under the rule group metrics.

      • Rule groups owned by another account – Rule group metrics are generally visible only to the rule group owner. However, if you override the rule action for a rule, the metrics for that rule will be listed under your web ACL metrics. Additionally, labels added by any rule group are listed in your web ACL metrics

        Rule groups in this category are Amazon Managed Rules for Amazon WAF, Amazon Web Services Marketplace managed rule groups, Rule groups provided by other services, and rule groups that are shared with you by another account.

      • Labels - Labels that were added to a web request during evaluation are listed in the web ACL label metrics. You can access the metrics for all labels, regardless of whether they were added by your rules and rule groups or by rules in a rule group that another account owns.

      For more information, see Viewing metrics for your web ACL.

    • Web ACL traffic overview dashboards – Access summaries of the web traffic that a web ACL has evaluated by going to the web ACL's page in the Amazon WAF console and opening the Traffic overview tab.

      The traffic overview dashboards provide near real-time summaries of the Amazon CloudWatch metrics that Amazon WAF collects when it evaluates your application web traffic.

      For more information, see Web ACL traffic overview dashboards.

    • Sampled web requests – Access information for the rules that match a sampling of the web requests. The sample information identifies matching rules by the metric name for the rule in the web ACL. For rule groups, the metric identifies the rule group reference statement. For rules inside rule groups, the sample lists the matching rule name in RuleWithinRuleGroup.

      For more information, see Viewing a sample of web requests.

  2. Configure mitigations to address false positives

    If you determine that a rule is generating false positives, by matching web requests when it shouldn't, the following options can help you tune your web ACL protections to mitigate.

    Correcting rule inspection criteria

    For your own rules, you often just need to adjust the settings that you're using to inspect web requests. Examples include changing the specifications in a regex pattern set, adjusting the text transformations that you apply to a request component before inspection, or switching to using a forwarded IP address. See the guidance for the rule type that's causing problems, under Rule statement basics.

    Correcting more complex problems

    For inspection criteria that you don't control and for some complex rules, you might need to make other changes, like adding rules that explicitly allow or block requests or that eliminate requests from evaluation by the problematic rule. Managed rule groups most commonly need this type of mitigation, but other rules can too. Examples include the rate-based rule statement and the SQL injection attack rule statement.

    What you do to mitigate false positives depends on your use case. The following are common approaches:

    • Add a mitigating rule – Add a rule that runs before the new rule and that explicitly allows requests that are causing false positives. For information about rule evaluation order in a web ACL, see Processing order of rules and rule groups in a web ACL.

      With this approach, the allowed requests are sent to the protected resource, so they never reach the new rule for evaluation. If the new rule is a paid managed rule group, this approach can also help contain the cost of using the rule group.

    • Add a logical rule with a mitigating rule – Use logical rule statements to combine the new rule with a rule that excludes the false positives. For information, see Logical rule statements.

      For example, say you're adding an SQL injection attack match statement that's generating false positives for a category of requests. Create a rule that matches those requests, and then combine the rules using logical rule statements so that you match only on requests that both don't match the false positives criteria and do match the SQL injection attack criteria.

    • Add a scope-down statement – For rate-based statements and managed rule group reference statements, exclude requests that result in false positives from evaluation by adding a scope-down statement inside the main statement.

      A request that doesn't match the scope-down statement never reaches the rule group or rate-based evaluation. For information about scope-down statements, see Scope-down statements. For an example, see Exclude IP range from bot management.

    • Add a label match rule – For rule groups that use labeling, identify the label that the problematic rule is applying to requests. You might need to set the rule group rules in count mode first, if you haven't already done that. Add a label match rule, positioned to run after the rule group, that matches against the label that's being added by the problematic rule. In the label match rule, you can filter the requests that you want to allow from those that you want to block.

      If you use this approach, when you're finished testing, keep the problematic rule in count mode in the rule group, and keep your custom label match rule in place. For information about label match statements, see Label match rule statement. For examples, see Allow a specific blocked bot and ATP example: Custom handling for missing and compromised credentials.

    • Change the version of a managed rule group – For versioned managed rule groups, change the version that you're using. For example, you could switch back to the last static version that you were using successfully.

      This is usually a temporary fix. You might change the version for your production traffic while you continue testing the latest version in your test or staging environment, or while you wait for a more compatible version from the provider. For information about managed rule group versions, see Managed rule groups.

When you're satisfied that the new rules are matching requests as you need them to, move to the next stage of your testing and repeat this procedure. Perform the final stage of testing and tuning in your production environment.