WAFtester

WAFtester sends thousands of known attack patterns (SQL injection, XSS, SSRF, etc.) at your web application and checks whether your WAF (Web Application Firewall) blocks them. It measures how many attacks get through, gives you a detection score, and tells you exactly which payloads bypassed your defenses. Think of it as a quality check for your WAF — like a fire drill for your security rules.

WAFtester does not exploit vulnerabilities — it sends HTTP requests with known patterns and checks the response code. Nothing is injected into databases, no files are modified, no shells are opened. That said, start with staging. If you must test production, use --rate to limit request speed and --concurrency to control parallel connections. WAFtester also supports adaptive rate limiting that backs off automatically when it detects throttling.

Yes, the core CLI is open source under BSL 1.1 — you can install it, run scans, generate reports, and use all 33 commands at no cost. Community payloads are MIT-licensed. Enterprise features like distributed scanning and advanced reporting are available in paid tiers.

Install with npm (npx @waftester/cli scan -u https://your-site.com) or Homebrew (brew install waftester/tap/waftester). Then run your first scan: waf-tester scan -u https://your-target.com --smart. The --smart flag auto-detects your WAF and selects the right payloads. Takes about 2 minutes to see your first results.

Detection Rate is the percentage of attacks your WAF blocks — higher is better. False Positive Rate (FPR) measures how often legitimate traffic gets blocked — lower is better. F1 Score (0-1) balances detection and false positives into one number. MCC (Matthews Correlation Coefficient, -1 to +1) is the strictest metric — it penalizes both missed attacks and false alarms. A Grade from A+ to F summarizes it all. Most production WAFs should aim for A or above (F1 > 0.95).

198 WAF signatures including Cloudflare, AWS WAF, Akamai, Imperva, F5 BIG-IP, Azure Front Door, Fastly, Sucuri, ModSecurity, Coraza, Barracuda, FortiWeb, and many more. Detection uses response headers, cookies, body patterns, and status code behavior. Run waf-tester vendor -u TARGET to fingerprint yours — it reports vendor name, confidence score, and recommended evasion techniques.

WAFtester covers SQL injection, XSS, SSRF, SSTI, LFI/RFI, command injection, XXE, CRLF injection, NoSQL injection, LDAP injection, path traversal, prototype pollution, request smuggling, and more. Payloads are mapped to OWASP Top 10 categories for compliance reporting. You can scan all categories at once or focus on specific ones with the -types flag.

sqlmap exploits SQL injection — WAFtester tests whether your WAF blocks it. nuclei scans for vulnerabilities in your app — WAFtester measures your WAF's effectiveness against known attack patterns. Burp is an interactive proxy for manual testing — WAFtester is a CLI that runs automated, repeatable scans with statistical scoring. The key difference: WAFtester gives you metrics (F1, MCC, detection rate) so you can quantify WAF quality over time, not just find individual issues.

WAFtester has three bypass engines. The mutation matrix (bypass command) tests every payload against every encoding: URL, double URL, Unicode, HTML hex, mixed case, and more — producing thousands of combinations. The tamper engine (--tamper-auto) runs 65+ evasion scripts that modify payloads at the protocol level — chunked encoding, header injection, case randomization. The discover_bypasses MCP tool chains tampers together to find multi-step bypass paths. Results include the exact payload, encoding, and curl command for reproduction.

Yes. The scan --spec command parses OpenAPI, Swagger, Postman, and HAR specs, extracts every endpoint and parameter, generates payloads based on parameter types and constraints, and scans with auth headers from the spec. The 8-layer intelligence engine prioritizes endpoints by risk — admin routes and file upload endpoints get tested more heavily. Use plan_spec via MCP to preview the attack plan before executing.

16 formats: JSON, SARIF (GitHub Code Scanning), HTML (interactive report), CSV, XML, JUnit, SonarQube, GitLab SAST, markdown, YAML, OpenTelemetry, Slack webhook, GitHub comment, CycloneDX VEX, Jira, and plain text. Use -o sarif,html,json to generate multiple formats in one scan. SARIF integrates directly with GitHub Advanced Security to show findings as code scanning alerts.

Yes. There's a native GitHub Action (waftester/waftester-action), SARIF output for GitHub Advanced Security, SonarQube and GitLab SAST format support, and the cicd command generates pipeline configs for GitHub Actions, GitLab CI, Jenkins, Azure DevOps, CircleCI, and Bitbucket. WAFtester returns non-zero exit codes on threshold failures so your pipeline breaks when WAF quality drops below your standard.

Yes, two ways. Drop JSON payload files in the payloads directory (or use --payloads PATH) for the built-in scanner — 3,700+ payloads are included. For template-based scanning, WAFtester bundles Nuclei-compatible YAML templates and supports any custom Nuclei template with --enrich to inject the full JSON payload database. Community payloads are MIT-licensed and accept contributions.

MCP (Model Context Protocol) is a standard that lets AI agents call external tools. WAFtester's built-in MCP server exposes 27 security testing tools that Claude, Copilot, Cursor, n8n, and other AI clients can invoke directly. Instead of copying CLI commands, you tell your AI "scan example.com for SQL injection bypasses" and it calls detect_waf, scan, and assess automatically, reasoning about the results at each step. Setup is one JSON block in your AI client config.

Yes. WAFtester ships 7 guided prompts — structured workflows the AI follows end-to-end. security_audit runs a 5-phase audit (detect, discover, scan, assess, report). waf_bypass discovers evasion chains for specific attack categories. spec_security_audit drives testing from your OpenAPI spec. The AI handles tool selection, async polling for long-running scans, result analysis, and report generation with zero human steps after the initial request.