Skip to content

Deep Dive — OWASP API Top 10 on Apigee

A gateway is a powerful control point, but it cannot fix everything. This page maps the OWASP API Security Top 10 (2023) to what Apigee can do — and, just as important, what must be done in the backend.

By the end you can...

  • Match each OWASP API risk to an Apigee control.
  • Recognise the risks a gateway cannot solve alone.
# Risk Apigee control Where in this guide Gateway enough?
API1 Broken object-level authorization (BOLA) Pass the verified identity (developer.email, claims) to the backend; never trust IDs from the client alone Lab 3 No — backend must check ownership
API2 Broken authentication VerifyAPIKey, OAuthV2, VerifyJWT (pinned algorithm, short expiry), no continueOnError on gates Labs 2–3 Mostly
API3 Broken object property-level authorization OASValidation with additionalProperties:false; strip fields in responses (AssignMessage/JS) OpenAPI-first Partly
API4 Unrestricted resource consumption SpikeArrest, Quota, ConcurrentRateLimit, payload-size & pagination caps, timeouts Lab 4, Access & routing Largely
API5 Broken function-level authorization Scopes per operation, products limited to paths/verbs OAuth grants Largely
API6 Unrestricted access to sensitive business flows Per-app/per-user quotas on specific flows, anti-automation (Advanced API Security, Cloud Armor) Platform Partly
API7 Server-side request forgery (SSRF) Never build target.url/callout URL from client input; allow-list targets Access & routing Largely
API8 Security misconfiguration apigeelint, TLS verification, normalised errors (no stack traces), CORS allow-list, encrypted KVMs Advanced Largely
API9 Improper inventory management One place for all APIs: products, API hub, versioning/deprecation (Sunset headers), analytics on old versions Advanced Partly
API10 Unsafe consumption of APIs Verify and size-limit responses from third parties (threat protection, OASValidation on responses), timeouts on callouts Lab 5 Partly

Worked example: hardening GET /orders/{id}

Weakest version: key check only. Stronger, in flow order:

  1. AccessControl (if internal) → blocks outsiders cheaply.
  2. SpikeArrest → stops floods.
  3. VerifyAPIKey / VerifyAccessToken → authenticates (API2).
  4. Scope or product path check → function-level auth (API5).
  5. Quota per app → consumption (API4).
  6. OASValidation → rejects malformed input (API3).
  7. Pass X-Authenticated-User to the backend; the backend confirms this user may read this order (API1).
  8. Response shaping removes fields the caller shouldn't see (API3).
  9. FaultRules return uniform errors with no internals (API8).

The gateway is not the last line

Anything depending on which object or which user needs backend checks. Treat Apigee as a strong first layer, not a substitute.

Experiments

  1. Take the labs bundle and list which of steps 1–9 it already has.
  2. Add the missing ones for /orders/{id} and re-run the smoke test.
  3. Write a one-line threat for each OWASP item as it applies to your API.

Check yourself

1. Why can't Apigee alone prevent BOLA?

BOLA is about whether this user may access this object; that data lives in the backend. The gateway can authenticate and forward identity but doesn't know ownership.

2. Which two policies most directly address unrestricted resource consumption?

SpikeArrest (rate) and Quota (allowance), plus size/pagination limits and timeouts.

3. How does a target.url built from a request header become SSRF?

An attacker supplies an internal address; your proxy then fetches it from inside the network. Resolve targets from an allow-list instead.