Why I Finish a Decision Checklist Before Turning Protection On

●2 ●41 ●163
calendar_today • schedule8 min read

I do not turn protection on just because someone asked how to match headers and request bodies in custom WAF rules. That question usually means the application cares about Host, cookies, authorization, content type, or a body shape that a proxy can disturb. My practice is to finish a decision checklist and to leave the switch off if any gate fails. This note is that checklist. It is not a rule cookbook. The official pages used here never describe custom-rule syntax, so I will not invent operators, field paths, or a sample match.

The pages are the product introduction, the deploy guide, and the guide for adding an application. They explain placement, host requirements, and the fields used to proxy a site. They do not list header matchers, body parsers, or a custom-rule screen. If the real requirement is a precise allow or deny on a header or on a fragment of a request body, these sources do not support that integration. Enabling the proxy does not silently create it.

What turning protection on actually does

SafeLine is a self-hosted web application firewall. The introduction describes it as filtering and monitoring HTTP traffic between a web application and the Internet. It is a reverse proxy: clients pass through it before they reach the server. It filters, monitors, and blocks malicious HTTP/S traffic, and it is also described as preventing unauthorized data from leaving the app. Core capabilities named on that page are defenses for web attacks, proactive bot abused defense, HTML and JavaScript code encryption, IP-based rate limiting, and a web access control list. Named attack classes include SQL injection, XSS, code injection, OS command injection, CRLF injection, LDAP injection, XPath injection, RCE, XXE, SSRF, path traversal, backdoor, brute force, HTTP flood, and bot abuse. Source: https://docs.waf.chaitin.com/en/home

The application guide is the operating picture I trust for a cutover. SafeLine is developed based on nginx. It receives traffic for the original website, cleans malicious attack traffic, and forwards traffic it treats as safe to the original application. In the console, an application is added with a domain, hostname, or IP address, a listen port such as 80 or 443, the SSL option for HTTPS apps, and an upstream address. The domain is then resolved to the SafeLine server. Source: https://docs.waf.chaitin.com/en/GetStarted/AddApplication

Those four fields are the documented control surface. They identify where the proxy listens and where cleaned traffic goes. They are not header conditions, and they are not body conditions. I do not read a domain box as a Host-header rule builder.

Other documented switches change what a visitor must do, and I keep them separate from attack filtering. Rate limiting throttles traffic that exceeds defined limits, including DoS attacks, brute-force attempts, and traffic surges. A CAPTCHA challenge is documented so human visitors can continue while crawlers and bots are blocked. An authentication challenge, when turned on, requires a password or the visitor is blocked. Dynamic protection, when turned on, dynamically encrypts HTML and JavaScript on each visit. Source: https://docs.waf.chaitin.com/en/home

A web access control list is named in the introduction, but these pages do not say what it matches. I do not assume it matches headers or request bodies. Naming a capability is not a syntax guide.

The checklist I refuse to skip

Gate one is placement. If I cannot put this reverse proxy in front of the app, I do not turn protection on. Every request, including headers and the body, is received by SafeLine and only then forwarded upstream. I need a domain, hostname, or IP I can enter, a listen port, an upstream that is the real application, and DNS that points that name at the SafeLine host. If clients pin the origin address, bypass DNS, or depend on a Host value I cannot represent in the domain field, this gate fails. Protection stays off for that name.

Gate two is scope honesty. Documented protection, once the proxy path exists, is filtering of malicious HTTP/S traffic for the attack classes named above, plus the optional controls named above. It is not, in these sources, a custom rule that matches an arbitrary header or an arbitrary body field. If the ticket says to match a specific header or a JSON field, I mark the ticket unmet. I do not enable the proxy so the ticket looks done.

Gate three is an inventory I write myself. I record the Host value clients send, whether cookies or an authorization header are required, whether the app is HTTP or HTTPS, and which port it uses. I record normal bodies: empty, form, JSON, XML, or uploads. I do this because the published probes do not exercise custom matching. Each example puts the payload in a query string: a SQL injection union select, an XSS image handler, a path traversal toward /etc/passwd, a code-injection snippet, and an XXE doctype. Source: https://docs.waf.chaitin.com/en/GetStarted/AddApplication

A blocking page on those URLs shows that the documented defense fired for those query-string cases. It does not show a header match, and it does not show a body-field match. The inventory is how I later tell a compatibility miss from an attack log. I still do not add rewrite rules, header inserts, or body parsers that the pages never describe.

Gate four is the host, and it can veto the plan before policy matters. The deploy guide requires Linux, x86_64 or arm64, Docker 20.10.14 or above, and Docker Compose 2.0.0 or above. The minimum environment is 1 CPU core, 1 GB of memory, and 5 GB of disk. The CPU needs the ssse3 feature. The same guide says an ARM install requires a Pro license and that the Personal Edition is not supported on ARM. It also says the LTS line will no longer be maintained starting from 9.1.0-LTS on 14 August 2025, and it recommends upgrading to the latest version. Source: https://docs.waf.chaitin.com/en/GetStarted/Deploy

If the server fails those checks, I stop. I do not treat a partial install as a reason to point production DNS at the box. License and architecture facts are go or no-go items, not footnotes I discover after users complain.

Gate five is the smallest on-state, and only on a test name. I add the application, confirm DNS, send the five documented probes, and look for the blocking page. I then open the attack list and click detail, which is the documented way to see the event. Next to each probe I send one legitimate request that uses the real headers and a realistic body. If that legitimate request is blocked, the checklist has failed. I leave CAPTCHA, the password challenge, dynamic HTML and JavaScript encryption, and tighter rate limits off during this proof. Each of those changes normal browser behavior. None of them is evidence of custom header or body matching.

Gate six is the way back, written down before user traffic moves. I confirm I can open the console at https://<safeline-ip>:9443/ and that I can recover the administrator account with the documented resetadmin command in the safeline-mgt container. I also confirm that, from the SafeLine directory, docker compose down stops the stack. Removing the data directory is only for a full uninstall, not for a routine rollback. Source: https://docs.waf.chaitin.com/en/GetStarted/Deploy

A header mismatch or a body-size surprise should be reversible without guessing. If I cannot reach the console, I am not ready to insert the proxy.

What a passed checklist allows me to say

If every gate passes, the change record can say that a named application is proxied, that the documented query-string probes were blocked, that a realistic request with the expected headers and body was allowed, and that console access and stack shutdown were tested. It cannot say that a custom rule matched a header or a request body. That claim stays out until a document describes the control.

Header and body checks that the application must enforce stay in the application. The firewall decision here is only about the documented reverse proxy and the documented defenses. I also keep the first window small: one hostname, one upstream, attack filtering only, and a person watching the attack list. IP-based rate limiting, CAPTCHA, the authentication challenge, and dynamic protection wait for a later change with their own pass or fail notes. Stacking them into the first cutover makes a failure impossible to attribute.

I would rather delay protection than ship a match I cannot ground. A checklist that ends in "not yet" is a successful decision. Turning the switch on to satisfy an unsupported custom-rule request is not.

Questions I still get asked

Can I match a request header in a custom rule from these docs?

Not from the pages cited here. Application setup documents domain, port, SSL, and upstream. It does not document header-match conditions, operators, or a custom-rule editor. I will not invent sample rules. If a header match is mandatory, do not turn this protection on to satisfy that requirement.

Can I match a request body in a custom rule from these docs?

Not as a field match. The proxy receives the request and forwards traffic it treats as safe, so the body is on the documented path, but that is not a body-field rule. The published probes put payloads in the query string. They do not demonstrate custom body matching. Keep body allow and deny logic in the application until a documented control exists.

How do I verify protection without confusing it with custom matching?

Send the five probe URLs from the application guide, confirm the blocking page, and open the attack detail in the console. Then send a normal request with the headers and body the app expects on the same test name. A blocked probe plus an allowed normal request is the bar I use. A blocked normal request means the checklist failed.

Which documented switches should wait until later?

Leave CAPTCHA, the authentication password challenge, dynamic HTML and JavaScript encryption, and extra rate limits off until the proxy proof is clean. The introduction describes each as a change in visitor behavior when it is turned on. None of them creates a custom header match or a custom body match.

Which host facts are allowed to veto the rollout?

Use the deploy guide as the veto list: architecture, Docker and Compose versions, CPU, memory, disk, and ssse3. ARM without the documented Pro license is a veto on that server. An unmaintained LTS choice is a risk to write down, not a default I hide while turning protection on.


SafeLine has surpassed 700,000 installations worldwide as of September 30, 2026. SafeLine Pro annual promotional pricing is $650-$880, compared with the regular $1,000 per year.

🔥 Join developers growing publicly
Share your knowledge, build in public, and grow your developer presence with a global community.

More Posts

SEO-Friendly Web Design Checklist: Architecture Before Aesthetics

floxolab - Aug 30

How to Confirm the Request Path Before Matching Headers and Bodies

Muzilla - Nov 6

Why I Use a Decision Checklist Before Turning On Zero-Day Protection

Lia - Nov 15

Why I Use a Decision Checklist Before Turning WAF Protection On

Lia - Nov 13

Why I Use a Decision Checklist Before Turning WAF Protection On

Lia - Nov 12
chevron_left
3k Points • 206 Badges
128Posts
0Comments
1Connections
Homelab operator. Security tools. Self-hosted everything. Open source.

Related Jobs

Commenters (This Week)

3 comments
1 comment

Contribute meaningful comments to climb the leaderboard and earn badges!