Skip to main content
Vulnerability Database/CVE-2026-77517

CVE-2026-77517: MaxKB Authorization Bypass Vulnerability

CVE-2026-77517 is an authorization bypass flaw in MaxKB open-source AI assistant that lets workspace users access other users' knowledge bases. This post covers technical details, affected versions 2.0.0-2.10.2-lts, and mitigation.

Published:

CVE-2026-77517 Overview

CVE-2026-77517 is an Insecure Direct Object Reference (IDOR) vulnerability in MaxKB, an open-source AI assistant for enterprise deployments. The flaw affects versions 2.0.0 through 2.10.2-lts. Document and paragraph API routes authorize only the knowledge_id supplied in the request path. The backend then looks up the target Document or Paragraph by its own UUID without verifying that the object belongs to the referenced knowledge base. A workspace user who knows a victim document_id or paragraph_id can read or modify content in another user's knowledge base by pairing it with an attacker-owned knowledge_id. The issue is tracked under [CWE-639: Authorization Bypass Through User-Controlled Key].

Critical Impact

Authenticated workspace users can read and modify documents and paragraphs owned by other tenants, breaking knowledge-base isolation across the MaxKB deployment.

Affected Products

  • MaxKB version 2.0.0 through 2.10.2-lts
  • MaxKB self-hosted enterprise AI assistant deployments
  • Multi-tenant MaxKB workspaces sharing a single instance

Discovery Timeline

  • 2026-09-21 - CVE-2026-77517 published to NVD
  • 2026-09-22 - Last updated in NVD database

Technical Details for CVE-2026-77517

Vulnerability Analysis

MaxKB exposes REST routes for managing documents and paragraphs inside a knowledge base. The route pattern embeds a knowledge_id that the framework uses for permission checks. That check confirms only that the caller has access to the knowledge base identified in the URL path. It never validates that the object referenced by document_id or paragraph_id is a child of that knowledge base.

An attacker who owns any knowledge base can substitute a victim's document_id or paragraph_id while keeping the attacker's own knowledge_id in the path. The controller loads the object directly from the database by its UUID and returns or updates it, bypassing tenant isolation. The result is unauthorized read and write access to another user's knowledge base contents, which can poison AI assistant outputs, exfiltrate proprietary data, or corrupt indexed documents.

Root Cause

The root cause is missing object ownership verification during authorization. Access control validates the container identifier in the URL path but trusts the child identifier without cross-checking the parent-child relationship. This pattern maps directly to [CWE-639], where an object reference from user input is honored without confirming the caller is entitled to that specific resource.

Attack Vector

Exploitation requires network access to the MaxKB API and low-privilege authenticated credentials in any workspace. No user interaction is needed. The attacker must obtain a valid UUID for a victim Document or Paragraph. UUIDs may be leaked through logs, shared links, browser history, error messages, or prior legitimate collaboration. The attacker then issues authenticated requests to the document or paragraph endpoints, keeping their own knowledge_id in the path and injecting the victim UUID into the object identifier parameter.

See the GitHub Security Advisory GHSA-58cm-c5jq-96vf for technical details.

Detection Methods for CVE-2026-77517

Indicators of Compromise

  • API requests where the document_id or paragraph_id path parameter does not resolve to a child of the supplied knowledge_id
  • Successful reads or writes on document and paragraph routes from user accounts that do not own the target knowledge base
  • Unexpected modifications to knowledge base content immediately following requests from unrelated workspace users

Detection Strategies

  • Enable verbose access logging on MaxKB and join request logs against the database to flag mismatches between path knowledge_id and the actual owner of the referenced document or paragraph
  • Baseline normal per-user access patterns to knowledge bases and alert on cross-tenant object access
  • Review web application firewall (WAF) or reverse proxy logs for enumeration patterns against /document/ and /paragraph/ routes

Monitoring Recommendations

  • Forward MaxKB application logs and API gateway telemetry to a centralized analytics platform for correlation across users, workspaces, and knowledge bases
  • Track document and paragraph modification events with user attribution and alert on writes by non-owning accounts
  • Monitor for high-volume UUID enumeration or repeated 200 responses on object endpoints from a single low-privilege account

How to Mitigate CVE-2026-77517

Immediate Actions Required

  • Restrict MaxKB API access to trusted networks and authenticated users until a patched version is released
  • Audit existing workspace accounts and remove or downgrade untrusted users to limit potential exploitation
  • Review recent document and paragraph access logs for cross-tenant activity and investigate anomalies

Patch Information

No fixed version is available as of this review. Monitor the MaxKB GitHub Security Advisory GHSA-58cm-c5jq-96vf and the MaxKB release notes for an updated version beyond 2.10.2-lts that enforces parent-child ownership validation on document and paragraph routes.

Workarounds

  • Deploy a reverse proxy or API gateway rule that inspects each request and confirms the referenced document_id or paragraph_id belongs to the knowledge_id in the path before forwarding
  • Isolate sensitive knowledge bases in separate MaxKB instances so cross-tenant IDOR is not possible within a single deployment
  • Rotate or invalidate any leaked document and paragraph UUIDs where feasible, and avoid exposing them in shared links or client-side code

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

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.