UTC
Medium

@arikusi/deepseek-mcp-server: Missing Authentication on Self-Hosted HTTP MCP Endpoint

CVE-2026-55605 · GHSA-72f3-6w86-7rv3 · published 2026-08-25 · updated 2026-10-01

Severity
Medium
CVSS score
5.3
Chance of exploitation (EPSS)
0.52%
Fixed in
1.8.0
  • Exploitation Public proof of concept
  • Automatable Yes
  • Technical impact Partial
  • CISA SSVC
What it is

In self-hosted HTTP deployments, any host with network access to port 3000 can establish an authenticated-equivalent MCP session and invoke built-in MCP tools without supplying credentials. This was verified end-to-end for session establishment, tool enumeration, and execution of the local deepseeksessions tool. deepseekchat is exposed through the same unauthenticated MCP session, and its handler routes requests through the server-side DeepSeek client. That means a deployment using a valid server-side DEEPSEEKAPIKEY places billable DeepSeek operations behind an unauthenticated endpoint.

What to do

Upgrade @arikusi/deepseek-mcp-server to 1.8.0 or later. If you start it from an MCP config with npx or uvx, pin that version so an old cached copy is not used.

Require authentication in self-hosted HTTP mode before MCP session creation. At minimum, pass an authProvider to createMcpExpressApp at transport-http.ts:17 or place equivalent authentication middleware / a reverse proxy in front of /mcp so unauthenticated clients never reach the initialize branch: For deployments that only need local access, bind to 127.0.0.1 instead of 0.0.0.0 in both createMcpExpressApp and app.listen (transport-http.ts:17 and :107) as a defence-in-depth measure. Upstream docker-compose.yml currently publishes 3000:3000; changing that to 127.0.0.1:3000:3000 would reduce inadvertent exposure on multi-user or server hosts.

Affected
PackageRegistryAffectedFixed in
@arikusi/deepseek-mcp-servernpm>= 1.4.2, < 1.8.01.8.0
Also in @arikusi/deepseek-mcp-server
  1. High@arikusi/deepseek-mcp-server has an Authorization Bypass Through User-Controlled Key@arikusi/deepseek-mcp-server

The impact, patch and workaround text comes from the advisory, shortened (GitHub Advisory Database, CC BY 4.0). Proof-of-concept and reproduction sections and code are left out; the original advisory has them.

Sources: GitHub Advisory Database, reviewed and malware advisories (CC BY 4.0, https://github.com/advisories); advisories published by Cursor (https://github.com/cursor/cursor/security/advisories); advisories published by OpenClaw (https://github.com/openclaw/openclaw/security/advisories); advisories published by vLLM (https://github.com/vllm-project/vllm/security/advisories); advisories published by Open WebUI (https://github.com/open-webui/open-webui/security/advisories); advisories published by n8n (https://github.com/n8n-io/n8n/security/advisories); advisories published by Flowise (https://github.com/FlowiseAI/Flowise/security/advisories); advisories published by Langflow (https://github.com/langflow-ai/langflow/security/advisories); advisories published by the MCP Python SDK project (https://github.com/modelcontextprotocol/python-sdk/security/advisories); advisories published by MLflow (https://github.com/mlflow/mlflow/security/advisories); CISA Known Exploited Vulnerabilities catalog (CC0). Data as of 2026-11-22. ShieldMap describes how each problem works and how to fix it; it does not publish exploit code. Report a correction to [email protected].