Skip to content

PATTERN Cited by 1 source

Positive security model API validation

Intent

Neutralise the advantage of automated exploit generators (including frontier AI models) by only permitting traffic that matches a known-valid schema, rather than trying to enumerate all invalid traffic.

Problem

Negative security models (block known-bad) degrade as the attacker can generate novel payloads faster than defenders can write rules. The defender is always playing catch-up.

Solution

For each API endpoint, define what a valid request looks like — from an OpenAPI spec, GraphQL schema, or learned from observed production traffic. Reject everything that does not conform. The model doesn't need to anticipate attacks; it only needs to know what is legitimate.

Mechanics

  1. Import or discover API schema (manual upload or traffic-learning)
  2. At the edge, validate every request against the schema for that endpoint
  3. Non-conforming requests are dropped before reaching origin
  4. Schema is continuously updated as the API evolves

Trade-offs

  • Pro: fundamentally resilient to novel attacks — attacker must produce a valid request, not merely an unknown one
  • Pro: works regardless of attacker volume (thousands of variants all fail)
  • Con: requires accurate schema maintenance (stale schemas block valid traffic)
  • Con: doesn't protect against abuse within valid schemas (needs additional business-logic checks)

Production example

Cloudflare API Shield: defines valid request shape per endpoint (from API definition or learned from real traffic). Traffic that doesn't fit is dropped.

(Source: sources/2026-06-09-cloudflare-defend-against-frontier-cyber-models)

Seen in

Last updated · 542 distilled / 1,571 read