GHSA-6QHC-X826-342C
Vulnerability from github – Published: 2026-06-16 21:02 – Updated: 2026-07-20 21:12Summary
The Docker API server applied its SSRF destination check to the crawl target URL only, not to the proxy address. An unauthenticated request could supply a proxy pointing at an internal IP and route the browser through it, reaching internal services and cloud-metadata endpoints, while using a perfectly valid crawl URL. The Docker API is unauthenticated by default.
Affected paths
/crawl, /crawl/stream, and /crawl/job accept a browser_config (and crawler_config). The following all feed Chromium's egress and were unchecked:
- browser_config.proxy_config.server
- browser_config.proxy (deprecated field)
- crawler_config.proxy_config.server
- --proxy-server / --proxy-pac-url / --proxy-bypass-list / --host-resolver-rules flags in browser_config.extra_args
Attack
An attacker sends /crawl with a benign, validation-passing URL but a proxy_config.server pointing at an internal IP. Chromium routes all requests through that proxy. For plain-HTTP targets the proxy receives the full request and can return any content, which is then returned verbatim in the crawl result (results[0].html / cleaned_html / markdown). In a real deployment the proxy would be an attacker-controlled server pointing at cloud metadata (e.g. AWS IMDSv1 at 169.254.169.254) to retrieve IAM credential tokens.
Impact
Unauthenticated server-side request forgery to internal services and cloud-metadata endpoints, with the response returned to the attacker.
Fix
Every proxy destination is validated with the same global-routability check used for crawl URLs (reject any resolved address that is not is_global, including IPv6 transition forms) before the browser is constructed; proxy/DNS-redirecting flags are stripped from extra_args. A legitimate public proxy still works. Honors CRAWL4AI_ALLOW_INTERNAL_URLS.
Workarounds
- Upgrade to the patched version (0.8.9).
- Enable authentication (
CRAWL4AI_API_TOKEN). - Restrict the container's outbound network access (egress firewall / no metadata route).
Credits
Geo (geo-chen) - reported the proxy_config.server SSRF with a clear PoC.
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 0.8.8"
},
"package": {
"ecosystem": "PyPI",
"name": "crawl4ai"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.8.9"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-53755"
],
"database_specific": {
"cwe_ids": [
"CWE-918"
],
"github_reviewed": true,
"github_reviewed_at": "2026-06-16T21:02:55Z",
"nvd_published_at": "2026-06-23T19:17:07Z",
"severity": "HIGH"
},
"details": "### Summary\n\nThe Docker API server applied its SSRF destination check to the crawl target URL only, not to the proxy address. An unauthenticated request could supply a proxy pointing at an internal IP and route the browser through it, reaching internal services and cloud-metadata endpoints, while using a perfectly valid crawl URL. The Docker API is unauthenticated by default.\n\n### Affected paths\n\n`/crawl`, `/crawl/stream`, and `/crawl/job` accept a `browser_config` (and `crawler_config`). The following all feed Chromium\u0027s egress and were unchecked:\n- `browser_config.proxy_config.server`\n- `browser_config.proxy` (deprecated field)\n- `crawler_config.proxy_config.server`\n- `--proxy-server` / `--proxy-pac-url` / `--proxy-bypass-list` / `--host-resolver-rules` flags in `browser_config.extra_args`\n\n### Attack\n\nAn attacker sends `/crawl` with a benign, validation-passing URL but a `proxy_config.server` pointing at an internal IP. Chromium routes all requests through that proxy. For plain-HTTP targets the proxy receives the full request and can return any content, which is then returned verbatim in the crawl result (`results[0].html` / `cleaned_html` / `markdown`). In a real deployment the proxy would be an attacker-controlled server pointing at cloud metadata (e.g. AWS IMDSv1 at 169.254.169.254) to retrieve IAM credential tokens.\n\n### Impact\n\nUnauthenticated server-side request forgery to internal services and cloud-metadata endpoints, with the response returned to the attacker.\n\n### Fix\n\nEvery proxy destination is validated with the same global-routability check used for crawl URLs (reject any resolved address that is not `is_global`, including IPv6 transition forms) before the browser is constructed; proxy/DNS-redirecting flags are stripped from `extra_args`. A legitimate public proxy still works. Honors `CRAWL4AI_ALLOW_INTERNAL_URLS`.\n\n### Workarounds\n\n- Upgrade to the patched version (0.8.9).\n- Enable authentication (`CRAWL4AI_API_TOKEN`).\n- Restrict the container\u0027s outbound network access (egress firewall / no metadata route).\n\n### Credits\n\nGeo ([geo-chen](https://github.com/geo-chen)) - reported the proxy_config.server SSRF with a clear PoC.",
"id": "GHSA-6qhc-x826-342c",
"modified": "2026-07-20T21:12:12Z",
"published": "2026-06-16T21:02:55Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/unclecode/crawl4ai/security/advisories/GHSA-6qhc-x826-342c"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-53755"
},
{
"type": "WEB",
"url": "https://github.com/pypa/advisory-database/tree/main/vulns/crawl4ai/PYSEC-2026-588.yaml"
},
{
"type": "PACKAGE",
"url": "https://github.com/unclecode/crawl4ai"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:N/A:N",
"type": "CVSS_V3"
}
],
"summary": "Crawl4AI: SSRF via proxy settings in the Docker server bypasses the crawl-URL SSRF check"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.