CVE-2026-67425: Insecure Credential Forwarding in Flyto2 Core
Vulnerability ID: CVE-2026-67425
CVSS Score: 8.6
Published: 2026-07-30
An insecure credential forwarding vulnerability in Flyto2 Core prior to version 2.26.6 allows attackers to exfiltrate operator API keys. This occurs because the system forwards environment-derived API keys to user-controlled custom endpoints, bypassing SSRF guards designed only for private target validation.
TL;DR
Flyto2 Core incorrectly forwarded high-privilege, environment-derived API keys to caller-controlled endpoints. This allowed unauthenticated remote attackers to exfiltrate sensitive keys by specifying a public, attacker-controlled host as the base URL.
⚠️ Exploit Status: POC
Technical Details
- CWE ID: CWE-522, CWE-201
- Attack Vector: Network (AV:N)
- CVSS v3.1: 8.6 (High)
- EPSS Score: 0.00319
- Impact: Confidentiality High (C:H)
- Exploit Status: Proof-of-Concept / Verified Private
- KEV Status: Not Listed
Affected Systems
- Flyto2 Core (flytohub/flyto-core)
-
Flyto2 Core: < 2.26.6 (Fixed in:
2.26.6)
Code Analysis
Commit: d5f89d7
Fix insecure credential forwarding to user-controlled custom endpoints (GHSA-qq9q-xgm3-xv9g)
Exploit Details
- GitHub: Proof of concept workflow setup describing extraction of keys using custom base_url triggers.
Mitigation Strategies
- Upgrade to Flyto2 Core 2.26.6 or later.
- Configure the FLYTO_TRUSTED_LLM_HOSTS environment variable to whitelist permitted custom API gateways.
- Establish firewall rules to restrict outbound HTTP/HTTPS traffic to trusted LLM endpoints only.
Remediation Steps:
- Pull and deploy the latest version (2.26.6+) of Flyto2 Core.
- Audit current configuration files and system environments for exposed API keys.
- Define FLYTO_TRUSTED_LLM_HOSTS in your runner environments if you utilize self-hosted or proxy model endpoints.
References
- Insecure Credential Forwarding in Flyto2 Core
- Fix Insecure Credential Forwarding Commit
- Flyto2 Core Release Tag v2.26.6
- NVD - CVE-2026-67425
Read the full report for CVE-2026-67425 on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)