CrowdSec vs Tailscale

The feed carries only release candidates, roughly one every two to three months, and the WAF has been the centre of gravity across all of them: OpenAPI schema validation, request body size limits, arbitrary AND/OR mixing in rule conditions, more Coraza transformations exposed, RE2 enabled by default on Linux. The 1.8.0 candidate breaks that pattern by adding challenge-and-fingerprint bot detection, serving an interstitial page and evaluating the result against configured rules. The same candidate adds a dedicated Kubernetes datasource that reads logs from the apiserver directly, and HTTP helpers so parsers and scenarios can query external services. Two earlier candidates in the window are pure refactoring with no user-visible change.

CrowdSec started as a log-reading detection engine that handed decisions to external bouncers, and the WAF work has been steadily moving it into the request path. Bot detection completes that move: the product now generates its own signal by interrogating the client rather than only inferring from logs. The Kubernetes datasource and the expression-language HTTP helpers point the same way — fewer intermediaries between CrowdSec and both the telemetry and the enforcement point.

Expect the fingerprinting rules to become a shared, community-curated resource in the same way detection scenarios already are, since that is the pattern this project applies to every new signal it collects.