Stop attacks before they are known: making the Cloudflare WAF smarter

Introduced a machine learning-powered WAF attack scoring system that classifies requests for malicious payloads, complementing signature-based rules. This system uses a model trained on millions of true positive samples from the Cloudflare network to detect evasion, bypass, and new attack techniques. The scoring system runs inline to HTTP traffic and is optimized for SQL Injection, XSS, and RCE attacks, with scores ranging from 1 (malicious) to 99 (clean). The post also announces new Security Analytics for visualizing attack score distributions and exploring unmitigated attacks before rule deployment. It details how to use Security Analytics to identify and mitigate attack traffic, and highlights the availability of WAF Attack Scores in HTTP logs and WAF Custom Rules.