CVE case study
CVE-2026-77522: Authenticated Full-Read SSRF in the MaxKB Web-Document Import Crawler
MaxKB through 2.10.3-lts let any authenticated workspace user add a web knowledge document by URL, and the server-side import crawler fetched that URL with no scheme allow-list and no block on loopback, link-local, or private addresses, then returned the response body to the caller. A low-privileged user could make the server read internal services and cloud instance-metadata credentials, a full-read, non-blind SSRF.
- Weakness
- CWE-918
- Affected
- MaxKB 2.10.3-lts and earlier
- Remediation state
- No patched release at disclosure; validate the URL scheme and host, block loopback, link-local, and private ranges, and remove verify=False
- Advisory published
- 2 Sep 2026
Official vectorCVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:N
Why it matters
MaxKB lets a workspace user add a web knowledge document or web data source by URL, and the server then crawls that URL and stores the fetched page as the document. The crawler issued the outbound request with no allow-list on the URL scheme and no block on internal address ranges such as 127.0.0.1, 169.254.169.254, or private networks, and it disabled TLS verification. Because the fetched body is returned to the requesting user, the request is non-blind: whatever the server can reach, the user can read.
The document-create and web-source flow is gated only at ordinary workspace-user level, not admin, so a low-privileged authenticated user could point the import at the cloud instance-metadata endpoint and read back IAM role credentials, or at internal-only admin APIs and services for reconnaissance. On AWS, GCP, or Azure deployments where the metadata endpoint is reachable, that can escalate to full cloud-account compromise.
The public record scores the issue Medium at 4.3 for its limited direct integrity impact, but the practical reach of a full-read SSRF that returns internal and metadata responses is what makes it worth fixing. The recommended remediation is to allow only http and https, resolve and reject loopback, link-local, private, and reserved addresses, re-check after redirects to defeat DNS-rebind and redirect bypasses, and restore TLS verification, ideally behind an egress proxy for crawler traffic.