No IP, No Problem: Exfiltrating Data Behind IAP

A cloud security researcher demonstrates how Google Cloud Platform's Identity-Aware Proxy (IAP) can be covertly abused to exfiltrate sensitive data from internal applications without requiring public IP assignments or external network access [00:00]. The talk begins with an architectural breakdown of GCP IAP, explaining how the proxy acts as a centralized perimeter that verifies user identities, enforces IAM policies, and injects authenticated identity tokens into requests forwarded to backend App Engine, Compute Engine, or GKE workloads [01:15, 02:11].

The researcher details a vulnerability stemming from cross-origin resource sharing (CORS) handling, where unauthenticated HTTP OPTIONS preflight requests are permitted through IAP by design to allow browser preflight negotiations [08:56, 09:06]. By gaining access to application configuration files (such as app.yaml) via CI/CD pipelines or internal App Engine deployment roles, an insider or compromised account can dynamically embed internal secrets into the Access-Control-Allow-Origin response header [13:26, 13:48]. An external attacker can then issue unauthenticated OPTIONS requests to harvest the reflected secrets from outside the protected network [11:26]. The session concludes with detection strategies using Cloud Audit Logs to identify rapid bursts of App Engine createVersion API events [15:42] and recommendations for hardening deployment pipelines.

(AI summary — Gemini Flash)