Reverse Proxy WAF: filter attacks before they reach your origin

A reverse-proxy WAF puts a proxy in front of your origin server instead of software on it. SucuraGuard runs that proxy on our own anycast network — Toronto, Chicago, Frankfurt, and Singapore — terminating TLS, inspecting every Layer-7 request, and forwarding only clean traffic downstream. Attacks are stopped upstream, before they hit your server. Point one A-record, free to start.

Quick Answer

A reverse-proxy WAF is an architecture, not just a feature: the firewall runs as a proxy in front of your origin server, so it terminates TLS, inspects Layer-7 requests, and applies rules before traffic ever touches your infrastructure — as opposed to a plugin that runs on your origin after a request has already arrived, or an appliance you deploy and patch yourself. SucuraGuard implements this as an anycast reverse proxy on Sucura's AS398999 network across four PoPs. You register your hostname and origin in Nexus, point one A-record at our anycast VIP, and clean traffic is all that reaches your server. Free to start, plans from $20/mo.

4 PoPs
Anycast Proxy
L7
Inspected In-Line
AS398999
Network
$0
Free Tier

How a Reverse-Proxy WAF Works

"Reverse proxy" describes where the firewall physically sits relative to your traffic. Instead of your origin server accepting connections directly from the internet, the DNS record for your hostname points at a proxy — in this case, an anycast VIP on our network. Every request for that hostname arrives at the proxy first, not at your server.

The proxy does four things before your origin ever sees a byte of the request. First, it terminates TLS — the encrypted connection ends at the proxy, using a certificate we issue and renew for your hostname. Second, it inspects the decrypted Layer-7 request — method, path, headers, cookies, body, and connection-level signals like TLS and HTTP/2 fingerprints. Third, it evaluates that request against a managed rule set plus any custom rules you've written, alongside bot-scoring, challenge, and rate-limit logic. Fourth, only the requests that pass are forwarded to your origin — usually re-encrypted over a fresh connection from the proxy to your server.

The result is that your origin server's job shrinks to "answer requests the proxy already decided are clean." It never has to parse a malicious payload, absorb a scanner's request flood, or spend CPU cycles evaluating whether a request is an attack — because that evaluation already happened upstream, on infrastructure that isn't your server.

Where the Filtering Happens

1

Request Hits The Proxy

Your hostname resolves to our anycast VIP, so the connection lands at the nearest of four PoPs, not at your origin.

2

TLS Terminates

The encrypted session ends at the proxy using an auto-issued certificate, so the request can be read in the clear for inspection.

3

L7 Rules Evaluate It

Managed and custom rules, bot score, rate limits, and challenge logic all run against the request before any decision to forward it.

4

Clean Traffic Forwarded

Only requests that pass reach your origin over a connection the proxy opens — attacks and bad probes never get that far.

Reverse Proxy vs. Plugin vs. Appliance

There are three common ways to run a WAF. Where the filtering physically happens is the whole difference between them.

Plugin WAF

Runs as code inside your application — a WordPress plugin, for example — after the request has already arrived at your origin. It adds PHP and CPU load to the same server it's protecting, and a large or malformed request still costs your server resources before the plugin gets a chance to block it.

Appliance / Self-Hosted WAF

A hardware box or software instance you deploy yourself, in front of your servers. It can filter upstream of the origin, but you provision it, size its capacity, patch it, and scale it — the operational burden sits with you, not a vendor.

Reverse Proxy WAF (SucuraGuard)

Runs as a managed proxy on our anycast network, upstream of your origin and upstream of our own patching cycle. Filtering happens before your server ever sees the request, and the capacity, uptime, and rule updates are ours to run — you just point one A-record.

How SucuraGuard Implements the Reverse Proxy

Every feature below is live in the product today — nothing on this page is roadmap.

A
Anycast Proxy, Four PoPs
One anycast VIP - your hostname's A-record points here; requests route to the nearest of Toronto, Chicago, Frankfurt, or Singapore
Same network as DDoS scrubbing - the reverse proxy runs on AS398999, the network Sucura already operates for Layer 3/4 mitigation
F
Request-Matching Signals
Network identity - source IP, ASN, network owner, and geography (country/continent/region/city)
Connection fingerprinting - TLS client fingerprint (JA3/JA4) and HTTP/2 fingerprint, useful for spotting scripted clients that spoof a normal user-agent
Behavioral signals - bot score, rate-exceeded, plus request path and header matching
R
Git-Versioned Custom Rules
Full rule history - every custom rule change is versioned with diff and revert, so you can see exactly what changed and roll it back
Rule limits by tier - 5 on Free, 25 on Pro, 250 on Business, unlimited on Enterprise
M
Managed Rule Set
Curated categories - known-CVE exploits, vulnerability scanners, probes, and other bad or malicious request paths
Not a named framework - a Sucura-managed rule set, not the OWASP Core Rule Set and not a ModSecurity deployment — updated centrally, on by default
T
Auto-TLS at the Proxy
Certificates pre-issued - before your DNS even points at us, so there's no gap in coverage during cutover
Custom certificate upload - available on Business, if you need to bring your own certificate rather than an auto-issued one
O
Origin Protection
Origin allowlisting - lock your origin firewall to accept connections only from our proxy's outbound ranges
Mutual TLS origin pull (Business) - your origin requires our client certificate before it will answer, so the origin isn't reachable by traffic that bypasses the proxy

Pricing

Per account, unlimited domains on every tier. Pick one in the Nexus panel — the Free tier costs nothing to try.

Free
$0
Unlimited domains
5 custom rules
Managed rule set
Basic challenge
Basic edge caching
1-day analytics
Pro
$20/mo
Unlimited domains
25 custom rules
Bot per-signal tuning
Credential-stuffing defence
Adaptive baseline
L7 rate limiting
Advanced caching
CAPTCHA challenge
API access
7-day analytics
Business
$200/mo
Unlimited domains
250 custom rules
Mutual-TLS origin pull
Custom certificate upload
Username-level credential-stuffing tracking
30-day analytics
Enterprise
Custom
Unlimited domains
Unlimited custom rules
Extended (~13-month) analytics retention
Prices are in USD; CAD billing is available. No per-request or metered charges — pricing is a flat monthly fee per tier.
Start Free in Nexus →

Frequently Asked Questions

What is a reverse-proxy WAF?
A reverse-proxy WAF is a web application firewall that sits in front of your origin server as a proxy, rather than running as software on the server itself. Requests hit the proxy first — it terminates TLS, inspects the Layer-7 request, applies rules, and only forwards traffic it decides is clean to your origin. Attacks are filtered upstream, before they ever reach your infrastructure.
How is this different from a WordPress plugin WAF?
A plugin-based WAF executes on your origin server, after the request has already arrived — it inspects traffic inside your own PHP process, which adds CPU and memory load to the same server it's protecting, and a large or malformed request still consumes your server's resources before the plugin can act. A reverse-proxy WAF filters the request at the edge, before it ever reaches your origin, so blocked traffic never touches your server's CPU, database, or application code at all.
Do I run or patch anything?
No. The proxy, the rule engine, and the underlying infrastructure are managed at our edge — you don't install software, size a box, or apply security patches. You register your hostname and origin in the Nexus panel and point one A-record at our anycast VIP; everything upstream of your origin is ours to run and keep current.
Is this managed ModSecurity or OWASP CRS?
No, and we want to be precise about this: SucuraGuard runs its own managed rule set — covering known-CVE exploits, vulnerability scanners, probes, and other bad or malicious request paths — plus whatever custom rules you write yourself. It is not the named OWASP Core Rule Set (CRS) and not a ModSecurity deployment, and we don't claim either. Think of it as a Sucura-curated rule set you don't have to license or maintain, not a third-party framework running under the hood.
How does it protect my origin IP?
Once your hostname resolves to our anycast VIP, all normal traffic to that hostname flows through the proxy rather than hitting your origin directly. For stronger protection you can also allowlist only our proxy's outbound ranges at your origin firewall, and on the Business tier you can require authenticated (mutual TLS) origin pull, so your origin only accepts connections that present our client certificate.
Does it terminate TLS?
Yes. Certificates for your hostname are issued and renewed automatically at the proxy, including pre-issuance before your DNS even points at us, so there's no certificate gap during cutover. If you need your own certificate rather than an automatically issued one, custom certificate upload is available on the Business tier.
Is Sucura the same as Sucuri?
No. Sucura Networks is an independent Canadian network operator running our own autonomous system, AS398999. We are not affiliated with Sucuri (the website security company owned by GoDaddy) or with GoDaddy — the similar name is a coincidence.
Start Free in Nexus Contact Us

Related SucuraGuard Resources