Skip to main content
Vulnerability Database/CVE-2026-108719

CVE-2026-108719: LLMGateway SSRF Vulnerability

CVE-2026-108719 is a blind SSRF vulnerability in LLMGateway through version 1.20.0 affecting the video-generation callback_url extension. This post covers technical details, affected versions, impact, and mitigation.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2026-108719 Overview

CVE-2026-108719 is a blind Server-Side Request Forgery (SSRF) vulnerability [CWE-918] affecting LLMGateway through version 1.20.0. The flaw resides in the video-generation callback_url extension, where the worker process delivers webhook POST requests without validating the destination through an assertSafeWebhookTarget check. API key holders can supply loopback, private, or cloud-metadata URLs, causing the worker to issue requests to internal services reachable from its network. The issue is tracked under CWE-918 (Server-Side Request Forgery).

Critical Impact

Authenticated attackers with a valid API key can coerce LLMGateway workers into reaching internal hosts, including cloud metadata endpoints, potentially exposing internal services and credentials.

Affected Products

  • LLMGateway (theopenco/llmgateway) versions through 1.20.0
  • Deployments exposing the video-generation API with callback_url support
  • Worker components that invoke deliverWebhook without egress validation

Discovery Timeline

  • 2026-10-11 - CVE-2026-108719 published to NVD
  • 2026-10-16 - Last updated in NVD database

Technical Details for CVE-2026-108719

Vulnerability Analysis

LLMGateway accepts a user-controlled callback_url parameter when clients submit video-generation jobs through the gateway API. The worker service dequeues the job and invokes deliverWebhook to POST status updates to the supplied URL. The code path through apps/gateway/src/videos/videos.ts, apps/worker/src/services/video-jobs.ts, and apps/worker/src/worker.ts does not call assertSafeWebhookTarget before dispatching the HTTP request. As a result, the worker will connect to any address the attacker provides, including private RFC1918 ranges, 127.0.0.1, link-local addresses, and cloud metadata endpoints such as 169.254.169.254.

Because the attacker does not receive the HTTP response body, the vulnerability is classified as blind SSRF. Attackers can still infer internal network topology through timing, error signaling, and side effects on internal services that accept POST requests. In cloud deployments, reaching the Instance Metadata Service (IMDS) can expose temporary credentials assigned to the worker role when IMDSv1 is enabled.

Root Cause

The root cause is missing egress validation on the callback_url field. A dedicated assertSafeWebhookTarget helper exists in the codebase but is not invoked in the video job delivery path. User input flows directly from API submission to outbound HTTP request without allowlisting, DNS resolution checks, or private-address rejection.

Attack Vector

Exploitation requires a valid API key, giving the attacker low-privilege authenticated access. The attacker submits a video-generation request with callback_url pointing at an internal resource. When the worker processes the job, it issues an HTTP POST to that resource from inside the trusted network boundary. Linked technical analyses in the VulnCheck LLMGateway Advisory and the HackMD SSRF Vulnerability Analysis document the affected code locations in the GitHub LLMGateway Repository.

No verified proof-of-concept code is published. Refer to the vendor code sections for the exact unguarded delivery path.

Detection Methods for CVE-2026-108719

Indicators of Compromise

  • Outbound HTTP POST requests from LLMGateway worker hosts to RFC1918, loopback, or link-local addresses
  • Worker-originated connections to 169.254.169.254 or other cloud metadata endpoints
  • Video-generation API submissions where callback_url resolves to internal infrastructure
  • Anomalous spikes in failed webhook delivery attempts referencing non-public hostnames

Detection Strategies

  • Inspect application logs for callback_url values that resolve to private, loopback, or metadata ranges
  • Correlate worker egress traffic with job submission records to identify operator-supplied internal targets
  • Deploy an outbound proxy that logs and blocks worker requests to non-allowlisted destinations

Monitoring Recommendations

  • Alert on any worker HTTP request to 169.254.0.0/16, 127.0.0.0/8, 10.0.0.0/8, 172.16.0.0/12, or 192.168.0.0/16
  • Monitor IMDS access logs on cloud instances hosting LLMGateway workers for unexpected credential retrievals
  • Track API usage per key for abnormal volumes of video-generation requests with unusual callback hostnames

How to Mitigate CVE-2026-108719

Immediate Actions Required

  • Upgrade LLMGateway to a version later than 1.20.0 once a patched release is published by the maintainer
  • Rotate any API keys that may have been used by untrusted tenants against affected deployments
  • Enforce IMDSv2 with hop limit of 1 on cloud instances running LLMGateway workers to blunt metadata exfiltration
  • Place workers behind an egress proxy that denies connections to private and metadata address ranges

Patch Information

At the time of publication, the advisory identifies versions through 1.20.0 as affected. Monitor the GitHub LLMGateway Repository for a fixed release that adds assertSafeWebhookTarget to the video-generation delivery path. The VulnCheck LLMGateway Advisory will track the fix commit.

Workarounds

  • Disable the video-generation endpoint or reject requests containing a callback_url field until a patched version is deployed
  • Validate callback_url at an upstream API gateway, resolving the hostname and rejecting any address in private, loopback, link-local, or metadata ranges
  • Restrict worker network egress with host firewall rules or a VPC egress policy that permits only known public webhook destinations
  • Issue API keys only to trusted tenants and scope keys to limit access to video-generation functionality
bash
# Example egress restriction using iptables on the worker host
iptables -A OUTPUT -d 169.254.169.254 -j REJECT
iptables -A OUTPUT -d 127.0.0.0/8 -j REJECT
iptables -A OUTPUT -d 10.0.0.0/8 -j REJECT
iptables -A OUTPUT -d 172.16.0.0/12 -j REJECT
iptables -A OUTPUT -d 192.168.0.0/16 -j REJECT

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.