Self-hosted web application firewall

See what’s hitting your site. Choose where to block.

Run DBshieldX behind your existing nginx. See which requests trigger a rule, review the traffic your customers depend on, and turn on blocking one route at a time. You can restore an earlier policy if you need to.

We monitor one site for a week, then walk you through what the firewall would have blocked. You choose when blocking starts.

We inspected about 158,000 requests across 16 of our own production sites in the 7 days to 7 November 2026. How we measure

shop.example.com · policy v5 Illustration
POST /checkout
7 days in monitor · review queue settled
Ready to block
POST /login
Blocking · credential-stuffing rule on
Blocking
POST /webhooks/payments
Pinned to monitor · partner traffic
Monitor
Review queue
SQL injection · field: q · GET /search
Attack Expected
Switch POST /checkout to block Roll back to v4
Latest alert Security alert: actively exploited OpenSSH flaw (CVE-2026-123158) 10 November 2026 · Read what to do
Fits the stack you already run nginxDocker · systemdCrowdSecAuthelia 2FAMaxMind GeoLite2Webhooks · n8n

Sound familiar?

bug_report

“Our logs are full of probes for .env, wp-login and SQL injection.”

Why it keeps happening

In the 7 days to 7 November 2026, block mode would have stopped 159 distinct clients on our sites. Of those, 148 (93%) were absent from the CrowdSec blocklists we checked.

toggle_off

“We have a firewall, but it only logs. Last time we switched blocking on, it stopped real customers.”

Why it keeps happening

A rule that looks right on a login page can also flag a valid payment notification. Until you review both, switching on blocking puts orders at risk.

storage

“A customer search set off an alert. Which part looked like SQL injection?”

Why it keeps happening

A count of alerts alone leaves you guessing. You need to see which rule matched, which field triggered it and where the request was going.

What it costs to leave it

A data breach

If customer data is exposed, you may have to report it to the ICO within 72 hours and tell the people affected.

A firewall nobody trusts

One wrong block costs orders and goodwill. Then blocking gets switched off and stays off.

Lost client confidence

For an agency, one compromised client site puts the whole relationship at risk.

What changes

Today
With DBshieldX
TodayYou see attacks in the logs after they have happened.
With DBshieldXYou see what would be blocked, page by page, before you block anything.
TodaySwitching on blocking for the whole site could interrupt checkout.
With DBshieldXYou can block on login while checkout and payment webhooks stay in monitor mode.
TodayA rule stops a customer from completing an order.
With DBshieldXYou can restore an earlier policy while you review the rule.
TodayA new scanner can reach you before its address appears on a blocklist.
With DBshieldXRules assess request content and behaviour. Plans with CrowdSec can also send ban decisions when the integration is enabled.
TodayA SQL injection alert leaves you searching through logs.
With DBshieldXThe finding identifies the rule, route and request field to review.
TodayYour traffic goes through someone else's network.
With DBshieldXEverything runs on your own servers.

How it works

A proxy behind your nginx inspects incoming web requests before they reach your app. Start in monitor mode to see what each rule flags on routes such as login, search and checkout. On our own traffic, the proxy added 3.2 ms at the 95th percentile in October 2026.

Review the traffic. Choose a route. Keep a way back.

visibility

01 · Monitor

See what it would block

Every route starts in monitor mode. DBshieldX logs what it would have stopped and never touches real visitors.

rule

02 · Review

Decide in one click

Mark each group of findings as an attack or expected traffic. Expected traffic becomes a precise exclusion for that rule, field and route.

block

03 · Block

Switch on with evidence

A route can block once it has run 7 days with its queue settled. You get an alert on its first block, and any block later marked expected counts as a false positive.

history

04 · Roll back

Every change is a version

Restore an earlier policy from the dashboard. Connected proxies pick it up without a restart.

How the rollout works

What the firewall looks for

Rules you can read, for the attacks that actually arrive.

Each finding names its rule and the field it was found in, so you always know why something was flagged.

storage

SQL injection

Unions, stacked queries, blind and time-based

code

Cross-site scripting

Script, event-handler and URL vectors

terminal

Command injection

Including shell evasion tricks

folder_open

Path traversal

Encoded, doubled and Unicode forms

dns

Server-side request forgery

Internal and cloud-metadata addresses

data_object

Template injection

Common server template engines

schema

NoSQL and LDAP injection

Operator and filter abuse

smart_toy

Bad bots and scanners

Tool fingerprints and probe paths

password

Abusive behaviour

Credential stuffing, enumeration, error bursts

DBshieldX inspects application requests behind nginx. Keep your CDN in place for high-volume DDoS attacks.

See full coverage

SQL injection detection

See which request field triggered a SQL injection alert.

When a search or form submission matches a SQL injection rule, the finding shows the route, the field and the rule. You can review what a customer was trying to do before you decide whether that route should block.

This check runs on incoming web requests. Automatic monitoring of live PostgreSQL queries is planned.

Proof

We run DBshieldX on our own sites first.

158k

requests a week inspected across 16 sites on 4 servers

159 / 159

clients block mode would have stopped in a week were scanners or fuzzers

93%

of the 159 clients block mode would have stopped were absent from the CrowdSec blocklists we checked

0

blocks later marked as expected traffic on our first blocking route in its first 7.8 days

7 days to 7 November 2026, from DBshieldX's own event records. Latency: p95 added by the proxy on live traffic, October 2026. How we measure

Security & privacy

Your traffic stays on your servers.

Self-hosted

The proxy, engine and dashboard run on your own machines. No request is sent to us.

Detail only when you opt in

By default, events record the rule, route and score, not the payload. Turn on evidence excerpts per tenant, with secrets masked and every view logged.

Retention you set

Detail is kept 30 days by default (7 to 90). Client IDs are hashed with a salt per tenant.

Tenants kept apart

Every customer sees only their own sites, keys and data. Dashboard sign-in uses two-factor single sign-on.

Fails open

If the engine is slow or stopped, requests go through and the error is logged. Your site stays up.

Protected by DBshieldX

Our own production sites run behind DBshieldX. Report a vulnerability to security@dbshieldx.com, or read our disclosure policy.

Pricing

Start free in monitor mode. Pay when you block.

A free Monitor plan for one site, Starter at £29 a month and Business from £99 a month (ex VAT), up to an Agency plan with a separate account for each client.

Compare plans

From the blog

FAQ

Start in monitor mode and review the requests each rule flags. You decide when to switch a route to blocking and can restore an earlier policy if a rule interrupts customer traffic.

Keep your CDN and DDoS protection in place. DBshieldX runs behind nginx and checks incoming requests for attacks against your application.

The firewall bases blocking decisions on rules you can read. Each finding shows which rule matched so you can review the decision.

Nowhere outside your servers. The proxy, engine and dashboard run on your own machines. If we install or manage DBshieldX for you, we handle data only on your instructions, under a data processing agreement.

Run the containers, then point your site's nginx upstream at the DBshieldX proxy, keeping your app as a backup line. We walk you through it on the demo call.

If the inspection engine times out or is unavailable, the proxy forwards the request to your app and logs the error. You can also configure nginx to reach your app directly if the proxy is unavailable.

Find out what your firewall would block in a week.

We monitor one site for a week, then walk you through what the firewall would have blocked. You choose when blocking starts.