Back to all lessons
Awareness Lessons
yesterday

SSRF Flaw in AWS MCP Gateway Exposed Internal Networks and Cloud Credentials

Written by AI from the linked story. Not reviewed by an editor. How this works

A server-side request forgery (SSRF) vulnerability in AWS's mcp-gateway-registry allowed authenticated users to coerce the gateway into making internal requests on their behalf, potentially reaching cloud metadata services and harvesting attached IAM role credentials. This is a classic SSRF risk that is amplified in cloud environments where instance metadata endpoints (e.g., IMDSv1) can expose powerful credentials without additional authentication. The vulnerability persisted across a wide version range (1.0.3–1.34.0), indicating that SSRF protections were not part of the original design or review process. The incident underscores how a single unvalidated outbound request path in an internet-facing service can pivot into full cloud account compromise if roles are over-permissioned or metadata protections are absent.

Tactical Insight

Immediate actions

  • Upgrade all mcp-gateway-registry deployments to version 1.35.0 or later immediately.
  • Audit IAM roles attached to the gateway service and restrict permissions to the minimum required.
  • Enable IMDSv2 (token-required mode) on all EC2 instances to limit metadata endpoint exposure via SSRF.

Long-term improvements

  • Integrate SSRF-specific testing (e.g., blind SSRF probes against internal endpoints) into the SDLC for all internet-facing services.
  • Enforce allowlist-based egress controls on gateway components so they can only reach explicitly approved destinations.
  • Apply the principle of least privilege to all cloud roles and rotate credentials on a scheduled basis.

Detection measures

  • Monitor outbound requests from gateway services for unexpected destinations such as 169.254.169.254 or internal RFC-1918 addresses.
  • Alert on any retrieval of cloud credentials or IAM token requests originating from the gateway process.
  • Review gateway access logs regularly for authenticated users making unusual or high-volume internal requests.

Back this up with a control

All controls and vendors

Training helps people spot this. These controls make the attack fail even on the day someone does not.