Always-on detections: eliminating the WAF “log versus block” trade-off
Key Points
- Always-on detections separate detection from mitigation
- Detection metadata (confidence,categories,ref) enables safe rule creation
- Full-Transaction Detection analyzes request+response to reduce false positives
Summary
Attack Signature Detection introduces an "always-on" detection layer that runs signatures on every proxied request and attaches rich metadata (confidence, categories, Ref IDs) before any mitigation. Detections are separated from mitigation so teams get full visibility into which signatures matched without sacrificing protection or performance. Full-Transaction Detection (request+response correlation) is in development to further reduce false positives and surface attacks that only appear in the response.
Key Points
- Always-on framework: signatures execute on every request and populate cf.waf.signature.request.{confidence,categories,ref} for use in Security Analytics and Security Rules.
- Low-latency design: detections can run post-origin when no blocking rule exists; when you create a blocking rule, detection moves inline and may incur additional latency depending on traffic.
- Signature model: signatures use Ref IDs, category tags (e.g., SQLi, XSS, RCE, CVE), and confidence (High, Medium) to guide safe blocking decisions.
- Analytics-first onboarding: enable detection, review aggregated matches and time-series in Security Analytics, run what-if scenarios on historical traffic, and scope exceptions for endpoints that trigger false positives (e.g., CMS rich HTML inputs).
- Rule creation: build rules from detection metadata (confidence and categories) to block, challenge, or create exceptions; combine tags and request attributes for granular policies.
- Availability: Attack Signature Detection is available in Early Access; Full-Transaction Detection is under development and will be available to early registrants.
Practical steps for engineers
- Opt in to Attack Signature Detection to begin collecting signature metadata immediately.
- Use Security Analytics to identify high-volume signatures and verify whether malicious requests reached the origin.
- Start with blocking High-confidence signatures; use historical what-if tests before enabling Medium-confidence signatures broadly.
- Create scoped exceptions for legitimate endpoints that trigger signatures, rather than disabling protections globally.
- Monitor latency after moving detections inline for any blocking rules and tune rules per application traffic profile.
Where detection fields appear
- cf.waf.signature.request.confidence — aggregated confidence scores (analytics & security rules)
- cf.waf.signature.request.categories — aggregated attack categories (analytics & security rules)
- cf.waf.signature.request.ref — aggregated signature Ref IDs (up to 10) (analytics & security rules)