Network protection vs Website protection
Every domain gets Network protection (L3/L4). Website protection (L7) is an optional extra you turn on per domain.
Network protection (L3/L4, always on)
Network protection works on network packets and connections, so it does not need to read or decrypt your visitors' requests.
- Per-IP limits on new connections per second and open connections on HTTP and HTTPS (ports 80 and 443). See Scraper throttling.
- Separate limits for Search engines and AI crawlers, recognised by their published IP ranges. See Verified crawlers.
- Temporary bans for addresses that keep going over their limits.
- Your deny list, plus stillorigin's shared list of known bad addresses.
- Optional PROXY protocol on HTTPS, so your origin sees visitor IP addresses without Website protection.
- Pass-through by hostname. For a domain with Website protection off, HTTPS connections are sent to the right origin based on the hostname the visitor asked for. The sentry does not decrypt the traffic and does not need your SSL certificate.
Website protection (L7, opt-in per domain)
Turn Website protection on with the Website protection button on the Domains page, or by adding a domain from the origin page as Website protection (starts in log mode). Each sentry includes one Website protection licence, pooled across the account. Turning Website protection on uses an included place first, then a Website protection licence; turning it off returns a licence to your pool if one was used. Extra Website protection licences are added on Billing. If the account goes over its Website protection allowance, Website protection is turned off on the newest domains and the oldest keep Website protection. When there is no spare capacity, the Off button is greyed out.
New domains with Website protection start in log mode. The path traversal and private file rules, and the WAF once you've turned it on for the cluster, record what they would block and let those requests through, until an account owner switches the domain to blocking. The WordPress hardening rules, rate limits, the WordPress admin login lock if it's on, and Network protection still block. Domains that already had Website protection on before this release stay in blocking.
With Website protection on, the sentry reads each HTTP request and can:
- reject requests for a hostname that is not one of your domains;
- block path-traversal attempts (
../) and requests for secrets and backup files (.env,.git,.sql,.bak,debug.log); - check requests for common attacks, such as SQL injection and cross-site scripting, with the WAF (OWASP Core Rule Set), once you've turned it on for the cluster;
- apply a per-IP request-rate limit, set by the cluster's Website protection requests per second and Website protection burst (answered with HTTP 429);
- apply the WordPress rules on domains of type WordPress;
- ask visitors to complete a quick, automatic browser check before the page opens, with Captcha Protection;
- pass the visitor's real IP to your origin in the
X-Forwarded-ForandX-Real-IPheaders; - show a stillorigin page when the origin cannot be reached.
Every request the sentry rejects is explained in Log reasons explained.
Every Website protection domain also comes with owner access: your customer can see their site's protection at owners.stillorigin.app, and change the settings you allow, without a stillorigin account.
With Website protection on, HTTP is inspected. For HTTPS, the origin agent sends the site's existing certificate to your sentries, sealed so only they can read it. See HTTPS inspection. Until a certificate is ready, HTTPS is passed through unopened, and Website protection still applies to HTTP.
The Network protection rules still apply to a domain when Website protection is on.
How the Domains traffic columns change
With Website protection off, HTTPS visits count under Passthrough HTTPS / hr, and the requests columns show mostly plain HTTP on port 80. With Website protection on, the requests columns include the Website protection rules. In log mode, requests those rules would have blocked count as allowed, not blocked. When HTTPS is inspected, those requests count under Allowed/hr and Blocked/hr, exactly like HTTP, and Passthrough HTTPS / hr drops to about zero. See HTTPS inspection and Domain traffic columns.