Der Shop ist online

Wir bauen Workstation AI-Maschinen für Websites, KI-Agenten und LLM-Training – online konfigurieren und kaufen.

Händlerpreise erhalten
KI-Produkte
KI-LabsOpenAI-AgentenClaude-AgentenGrok-BotWorkstation AI-natives CRMMarketingAlle Produkte
KI-Lösungen
Angewandte KIKI-AgentenGenerative KIKI-AutomatisierungKI-WeiterbildungKI-Pakete für KMUEnterprise AI LabKI-WorkstationsNVIDIA-Hardware & -SoftwareNVIDIA Accelerated ComputingPrivate KIGPU-Private-CloudGPU-ClusterEdge-KIKI nach Branche
KI-Services
PlattformmodernisierungDigital EngineeringDatengrundlagen & KIAutonomer BetriebKI-BeratungDevOps-AutomatisierungCybersicherheitSoftwareentwicklungAgentenentwicklungMLOps-EinrichtungGPU-LeasingGPU-Private-Cloud
Über uns
UnternehmenUnsere GeschichtePartnerKundengeschichten
KI-Rechenzentren
KI-RechenzentrenGPU-Private-CloudGPU-ClusterGPU-Leasing
KI-EinblickeJev-ReiheBenchmark LabModellauswahl
Technische Hinweise
Dokumentation
Linux-BetriebssystemeUbuntuRed Hat Enterprise LinuxSUSE LinuxAlpine LinuxDebian, Rocky & AlmaLinux
KI-Cloud-PlattformenAWS & Amazon BedrockGoogle Cloud & Vertex AIMicrosoft Azure & AI FoundryGPU-Clouds & Marktplätze
Private KI-RechenzentrenFür AnwaltskanzleienFür Steuer- und WirtschaftsprüfungskanzleienOn-Premises-PlattformenVollständig abgesicherte Cloud
ToolsRing PromoterJobshout
News
Blog
KontaktAnmelden

KI-Workstations, Multi-Agenten-KI-Software, GPU-Infrastruktur und intelligente Agentenlösungen für moderne Unternehmen.

Kontakt

KI-Lösungen

Angewandte KIKI-AgentenGenerative KIKI-AutomatisierungKI-WeiterbildungKI-Pakete für KMUEnterprise AI LabKI nach Branche

Produkte

Alle ProdukteWorkstation AI-natives CRM & ERPMarketingOpenAI-AgentenRing PromoterJobshout

Unternehmen

Über unsUnternehmenKI-RechenzentrenWarum WorkstationPartnerKundengeschichtenPreiseKontaktTenthMatrix

Ressourcen

KI-EinblickeTechnische HinweiseDokumentationLinux-BetriebssystemeKI-Cloud-PlattformenPrivate KI-RechenzentrenBlogSucheSitemap
Büro Belgien
Workstation SRL, Rue Vanderkindere 34, 1180 Uccle, BrusselsBE 0751.518.683Mo – Fr: 9:00 – 18:00 Uhr MEZ
+32 492 45 67 46
Hauptsitz Großbritannien
77-79 Marlowes, Hemel Hempstead HP1 1LFAnfahrt – Ausfahrt 20 der M25 (Outer London)Handelsregisternr.: 11641870Mo – Fr: 9:00 – 18:00 Uhr GMT
+44 7515 356 146
Büro Indien
#159 Sector 9, Pocket 1, DDA Flats, 110077 Dwarka, New Delhi
+91 98881 98841
Büro Taiwan
3F, No. 2-4, Aisi Rd., Ren'ai Dist., Keelung City 200, Taiwan基隆市仁愛區愛四路2之4號3樓Mo – Fr: 9:00 – 18:00 Uhr Taiwan-Zeit (GMT+8)
+886 926 892 701+32 492 45 67 46
Büro Estland
Aleksander Puškini tn 30-25, 20306 Narva, EstoniaMo - Fr: 9:00 - 18:00 Uhr estnischer Zeit (EET)
+32 492 45 67 46

© 2026 Workstation AI. Alle Rechte vorbehalten.

DatenschutzDatenlöschungCookiesNutzungsbedingungenWebsite-Sitemap

NVIDIA, das NVIDIA-Logo, RTX, RTX PRO, Blackwell, CUDA, NVLink, MGX, Omniverse, Spectrum-X, Quantum, BlueField, DGX und HGX sind Marken und/oder eingetragene Marken der NVIDIA Corporation in den USA und anderen Ländern. AMD, das AMD-Arrow-Logo, EPYC, Instinct, Ryzen und Threadripper sind Marken von Advanced Micro Devices, Inc. Dell und PowerEdge sind Marken der Dell Inc. oder ihrer Tochtergesellschaften. Intel und Xeon sind Marken der Intel Corporation oder ihrer Tochtergesellschaften. AMAX, AceleMax und LiquidMax sind Marken der AMAX Engineering Corporation. Alle anderen Produktnamen, Logos und Marken sind Eigentum ihrer jeweiligen Inhaber. Wir verwenden sie ausschließlich, um die Produkte zu kennzeichnen, die wir beschreiben, liefern oder unterstützen; ihre Verwendung bedeutet nicht, dass ihre Inhaber uns empfehlen.

Loading blog...

Startseite / Blogs
DevOpsSecuritySREKubernetesAIAutomation

WSLProxy WAF Engine v2: Enterprise Enforcement You Can Explain

Stable signature IDs, monitor/block bindings, and support IDs on every decision

Balinder Walia17. September 20266 min read
Workstation WSL Proxy WAF Engine v2

Workstation WSL Proxy ships WAF Engine v2 as an explainable enforcement layer on the OpenResty hot path — not a regex snippet nobody owns. Stable signature IDs, monitor/block bindings, and a support ID on every decision. Deep dive: Inside WAF Engine v2. Product: wslproxy.org/waf · Overview: /wsl-proxy.

Workstation WSL Proxy WAF Engine v2

Watch: WSL Proxy WAF tour

Workstation WSL Proxy WAF tour

Watch: youtu.be/r10XSonA5JE · Product: wslproxy.org · WAF: wslproxy.org/waf

Bottom line. Opaque WAFs create 2 a.m. tickets with no policy name and no safe dry-run. Engine v2 is the implemented MVP on Workstation WSL Proxy: eight first-class stages, governed signatures, binding precedence you can log, and fail-open by default so a WAF bug cannot take the site down. Inspired by F5 WAF for NGINX (App Protect) goals — not a parity claim.

Opaque edge security vs explainable enforcement

Platform and security teams still inherit three failure modes at the edge. First: a static nginx snippet or ModSecurity pack that nobody owns, so every change is a POP-wide reload argument. Second: a cloud checkbox that blocks traffic without naming the winning rule, binding, or stage. Third: a “monitor mode” that is really just a different product with no path to promote the same policy to block per route.

When something breaks, the ticket says “403 from the edge.” There is no policy ID, no support ID, and no way to stage a noisy signature without rewriting the pack. That is not enterprise enforcement — it is folklore with a status code.

Workstation WSL Proxy treats the edge as a product. Day-to-day routing and WAF policy move on the live-config hot path; nginx reloads stay reserved for listen and SSL. Engine v2 is the security half of that story: a production WAF you can bind per domain and route, run in blocking or transparent mode, govern by stable signature IDs, and operate with structured logs plus correlation IDs.

What Engine v2 is (MVP, shipped)

Engine v2 runs in the OpenResty request path in three Lua modules: api/waf_engine.lua, api/waf_stages.lua, and api/waf_support.lua. It is the design reference for architecture, policy schema, binding precedence, signature governance, observability, and the roadmap that maps remaining App Protect–style features to Phase 2 and Phase 3 — clearly labelled future.

The goal is specific: a WAF that platform and security teams can bind, explain, and roll out safely — not a regex list pasted into a location block. On Workstation’s site we call the product Workstation WSL Proxy; on the product surface it is WSLProxy at wslproxy.org.

If you already read the broader WAF overview (WSL Proxy WAF blog), this piece focuses on Engine v2 itself: constraints, pipeline, operator levers, and where the roadmap stops being “done today.”

Four design constraints in plain language

Constraint Meaning for operators
In the NGINX pathLuaJIT, compiled-regex cache (o flag), shared dicts for velocity. No blocking I/O on the hot path.
Fail-open by defaultEngine or stage errors log and allow. A WAF bug must not take the site down. Fail-closed body parse is roadmap Phase 2 — not the default.
Backward compatiblev1 policies (waf_rules + mode) keep working; every v2 field is optional and additive.
ExplainableEvery block names policy, winning binding, stage, violation code, signature ID, and support ID.

Those four constraints are why Engine v2 feels different from a bolted-on rule file. Availability bias is intentional. Compatibility means you can adopt v2 fields without rewriting every host. Explainability is not a marketing adjective — it is what lands in response headers and wafsec logs.

The pipeline in sixty seconds

A request hits a POP. gateway_ack selects the route. gateway_pipeline runs rate-limit, then WAF, then the rest of the response path. Inside waf_engine.inspect (fail-open wrapper):

  1. Load policy (30s TTL cache).
  2. Resolve effective mode and winning binding.
  3. Run the eight-stage pipeline (pure functions: each returns a finding or nil).
  4. Run governed signature matching and anomaly threshold.
  5. On block: 403 plus X-WAF-Block, X-WAF-Rule, X-WAF-Violation, X-Support-ID, and the block page.

Stages do not decide block versus alarm. The engine does, from the effective enforcement mode, so the same rule set can behave differently per binding without duplication. Stages cover method allow-list, filetype deny, HTTP smuggling/desync, IP and geo, JWT algorithm policy, JSON depth/size, brute-force velocity, and OpenAPI path+method positive security. Signatures then match with disable / stage / set governance and optional anomaly scoring.

gateway_ack → gateway_pipeline (rate-limit → WAF → …)
  → waf_engine.inspect
    → load policy → resolve binding
    → STAGE PIPELINE (8) → SIGNATURES + anomaly
    → block: 403 + X-WAF-* + X-Support-ID

For the annotated schema, binding algorithm, and full violation catalogue, see the long technical article.

Why operators care: binding, staging, support IDs

Binding precedence is most-specific-wins: route override beats per-server waf_mode_override, which beats policy enforcementMode. Longest path prefix wins, so /api/admin beats /api. Every finding records binding — domain, server, or route:<path> — so logs answer which lever won. Today service is a logical label in logs; a service→policy selection map is Phase 2 (future).

Signature governance gives three independent controls: disable by ID (skip entirely), stage until a timestamp (alarm only; does not contribute to anomaly score), and set toggles (block:false downgrades a whole set to alarm-only). That is the safe rollout path: soak noisy signatures without risking an indirect block via the anomaly threshold.

Support IDs use the form WSL-<epoch>-<rand>, echoed as X-Support-ID and rendered into the block page. Structured wafsec JSON lines, the waf_events shared dict / recent-events API, and Prometheus counters (waf_blocked, waf_monitored, waf_inspections, waf_latency, waf_errors) give SRE and security the same correlation story.

Practically: bind a payments policy in blocking on /api/pay, leave /preview transparent, stage a new open-redirect signature until Friday, and when a customer pastes a support ID into a ticket you look it up — you do not guess which anonymous rule fired on another system.

That workflow is what “enterprise enforcement you can explain” means on Workstation WSL Proxy. Security and platform teams share one audit surface with routing and rate limits. Agents and humans can inspect the same policy objects through Admin UI, REST, MCP, or CLI without waiting for a POP-wide nginx reload for ordinary WAF changes.

Roadmap (labelled future — do not blur with MVP)

MVP is done today: domain and route bind, signature sets and per-ID disable/stage, method and filetype, IP and geo, JWT alg, JSON profile, brute-force, OpenAPI path+method, smuggling stage plus signature, structured logs and support IDs, Prometheus, block page, golden tests, CI validation, and Admin UI for WafPolicies/WafRules. Phase 2 (future) includes the service→policy map, deeper OpenAPI param/type checks, XML and GraphQL profiles, cookie integrity, JWT JWKS, Data Guard, references in events, fail-closed parse option, bot/JA3 signals, threat packs, and an aegisctl-style CLI. Phase 3 (future) covers behavioural L7 DoS, IP reputation, MaxMind geo, gRPC/protobuf, Ingress CRDs, and Hyperscan/Vectorscan. Non-goals for v1 remain full RASP, proprietary signature DB clones, and in-path ML training.

Where to go next

  1. Watch the tour above, then open wslproxy.org/waf.
  2. Read Inside Workstation WSL Proxy WAF Engine v2 for stages, schema excerpts, binding pseudocode, and the full violation table.
  3. Skim the product overview at /wsl-proxy and the earlier WAF narrative at WSL Proxy WAF.
  4. Engine design reference in source (until a dedicated docs site exists): docs/WAF_ENGINE_V2.md on GitHub.

Docs website deferred — this blog and the companion article point at GitHub docs/WAF_ENGINE_V2.md and wslproxy.org/waf.

Published by Workstation.

Product site: https://wslproxy.org · WAF: https://wslproxy.org/waf/