API pentesting, where the real flaws live.
AI penetration testing for REST and GraphQL APIs, covering broken object-level authorization, mass assignment, query-depth abuse, authentication flaws, and business logic vulnerabilities.APIs are where business logic runs. We test REST and GraphQL surfaces for the flaws scanners consistently miss: broken object-level auth, mass assignment, query abuse, and the chained bugs that turn low-severity findings into takeover.
APIs are where scanners fail.
A scanner sees a fuzzed endpoint. An API pentester sees a stateful conversation with auth, logic, and side effects. The bugs that matter are the ones you can only find by reasoning about the API, not fuzzing it.
The OWASP API #1 for a reason
Broken object-level auth is the most common serious API flaw, and it's the hardest to automate. It requires knowing whose data is whose. The AI tests systematically with multiple accounts; humans verify the business impact.
A whole new surface, same old scanners
Query depth, alias abuse, introspection in prod, batching attacks, field-level auth bypass. GraphQL is a different protocol and needs different tradecraft. We've got it; most scanners don't.
Low-severity bugs, high-severity outcomes
An info disclosure plus a mass assignment plus a weak rate limit is a takeover. Scanners report each finding in isolation. Our AI chains them; our humans verify the chain actually works end to end.
What we test on APIs.
REST and GraphQL. Authenticated and unauthenticated. Internal and external. Every engagement.
Who can do what
- BOLA / IDOR across resources
- BFLA · admin methods on user roles
- Tenant isolation gaps
- Object-property auth flaws
- Scope escalation
- Service-to-service auth abuse
HTTP's own quirks
- Mass assignment
- HTTP method tampering
- Verb override abuse
- Parameter pollution
- Content-type confusion
- Rate-limit bypass (headers / IP)
Query language attacks
- Introspection in production
- Query depth / complexity DoS
- Alias-based rate-limit bypass
- Batched query abuse
- Field-level auth gaps
- Mutation via query chaining
What the API actually does
- Business-logic abuse
- Race conditions
- Replay attacks
- Data exfiltration via pagination
- Batch endpoint abuse
- Webhook SSRF
Example findings from this surface.
Illustrative examples of real vulnerability classes on this attack surface. Anonymized, but every one based on patterns our pentesters encounter repeatedly.
The invoice endpoint trusted the path-level user ID without checking it matched the authenticated session. Any valid session could enumerate every user's invoices. 120k records exposed in testing scope.
The full schema was queryable in production via `__schema` introspection. This revealed internal admin mutations that weren't meant to be exposed. One of them had a BFLA flaw.
The profile update endpoint deserialized the request body into the User model directly. Including `"role": "admin"` in the JSON body elevated the user to admin. No audit log fired.
Login rate-limiting was enforced per HTTP request. A single GraphQL query could include 100 aliased login mutations, each trying a different password. Effective rate: 100x the limit.
// illustrative examples · not real customer engagements
Ship faster than your pentesters? Let's fix that.
Tell us what you ship and we'll scope a pilot on this surface. The AI and the reviewer take it from there.
Request access