What feature of web security filters websites based on who what where, how, and when to assign a website a score?

Sandbox Filtering executes suspicious files or URLs in an isolated virtual environment to observe behavior - it does not assign scores based on contextual attributes of a website.

Reputation Filtering uses a multi-factor scoring model that evaluates who owns the domain, what content it serves, where it is hosted, how it behaves over time, and when it was registered to generate a numeric trust score. This score is then used by the web security platform to allow, block, or warn users before a connection is established. Cisco Web Security Appliance and similar products use this scoring as a primary enforcement mechanism.

Malware Filtering scans content for known malicious signatures or heuristic patterns, not a holistic scoring of websites using contextual who/what/where/how/when factors.

URL Analytics refers broadly to analyzing URL structure and patterns for categorization, but it is not the specific mechanism that combines multiple contextual signals to produce a reputation score.