Waf

AWS WAF

Overview

Domain expertise for configuring AWS WAF, the web application firewall that filters HTTP and HTTPS traffic to CloudFront distributions, Application Load Balancers, API Gateway REST APIs, and AppSync GraphQL APIs. Covers web ACL creation and association, AWS Managed Rules, rate-based rules, match rules (IP set and geographic), Bot Control and the signal-forwarding workflows built on top of it, Fraud Control for logins and signups, AI and LLM crawler management, and the logging that every tuning workflow depends on.

This skill is a router. Each customer task maps to a procedure file under references/. Read the matching reference in full before acting, then follow its constraints and steps. The reference files are self-contained: each carries its own decision tables, constraints, procedure, and troubleshooting.

Execute commands using the AWS MCP server when connected (sandboxed execution, audit logging, observability). Fall back to the AWS CLI otherwise. A web ACL’s scope is fixed at creation: a CloudFront web ACL must be created in us-east-1 with CLOUDFRONT scope, while a regional web ACL (Application Load Balancer, API Gateway, AppSync) is created in the resource’s Region with REGIONAL scope.

Which WAF task do you need?

Routing notes

• Logging comes before tuning. Every Count-mode tuning workflow assumes logging and request sampling are on. If the customer has not set up logging, run that reference first; otherwise Count-mode tuning has nothing to read.
• Web ACL scope is fixed at creation. A CloudFront web ACL is CLOUDFRONT scope in us-east-1; a regional resource needs a REGIONAL web ACL in its own Region. Scope cannot be changed later, so the creating reference settles it before anything is built.
• Bot Control is a chain, not one task. Protecting against bots is the on-ramp (turn on, choose Common vs Targeted, observe). Turning labels into a confidence signal, forwarding that signal, and deciding what the application does with it are three separate references that build on it in that order. The header-stripping reference is the mandatory safety companion whenever a signal is forwarded to the origin.
• Common vs Targeted is not a soft choice. Common only catches self-identifying bots and known-bad IPs. For login, checkout, or any high-value endpoint facing evasive bots, Targeted with the application integration SDK is required. The bots reference pushes Targeted for real bot threats rather than presenting it as optional.
• Rate limiting vs Fraud Control. Rate-based rules blunt volumetric HTTP floods. Credential stuffing and fake-account creation are account-based abuse that rate limiting misses; those go to the Fraud Control reference (ATP and ACFP), not the rate-based reference.
• Forwarded headers need the strip rule. Any time the customer forwards a signal or the client IP to the origin in x-amzn-waf-* headers, the inbound-header-stripping reference is required to prevent spoofing. The confidence-signal, interpolation, and client-IP references all point at it.
• What lives in other skills. L3/L4 DDoS protection and Shield cost-protection credits are the shieldadvanced skill. Multi-account WAF rollout is the firewallmanager skill. CloudFront and Application Load Balancer configuration are their own skills. This skill builds the WAF rules; it does not configure the resources it protects.

Security Considerations

AWS WAF is itself a security control, so misconfiguration directly weakens an application’s defenses. Apply these across every reference:

• Least-privilege IAM. You MUST grant only the specific wafv2: actions a task needs (for example wafv2:CreateWebACL, wafv2:GetWebACL, wafv2:UpdateWebACL, wafv2:AssociateWebACL, wafv2:PutLoggingConfiguration) rather than wafv2:* or the AWSWAFFullAccess managed policy.
• Ephemeral credentials. You MUST use IAM roles with temporary credentials (such as an EC2 instance profile, SSO session, or aws sts assume-role) rather than long-lived IAM user access keys when running these WAF CLI commands.
• Monitor configuration changes. You SHOULD enable AWS CloudTrail on wafv2 management events and set CloudWatch alarms on critical web ACL configuration changes (such as DeleteWebACL and UpdateWebACL rule removals) and on the web ACL’s BlockedRequests and CountedRequests metrics, so rule changes and sudden spikes in blocked or counted traffic are detected.
• Misconfiguration opens access. A web ACL that is created but never associated, or one whose default action is left at Allow with no enforcing rules, filters nothing. You MUST confirm the web ACL is associated and that its posture matches the intended default (block vs allow) before reporting setup complete.
• Protect log destinations. Logs can capture credentials and session data. You MUST redact sensitive fields (such as the authorization header and cookie) and MUST enable encryption at rest on the log destination (CloudWatch Logs, Amazon S3, or Amazon Data Firehose).
• Header-spoofing risk. Any x-amzn-waf-* signal forwarded to the origin can be forged inbound. You MUST add the inbound-header-stripping rule whenever a signal or client IP is forwarded (see stripping-inbound-waf-headers-before-trusting-them).

Additional Resources

• AWS WAF Developer Guide (opens in a new tab)
• How AWS WAF works (AWS WAF Developer Guide) (opens in a new tab)
• AWS WAF pricing (opens in a new tab)