Skip to main content
CVE Vulnerability Database
Vulnerability Database/CVE-2026-68578

CVE-2026-68578: ArcadeDB Auth Bypass Vulnerability

CVE-2026-68578 is an authentication bypass flaw in ArcadeDB that allows non-root users to perform arbitrary database operations and execute JavaScript code. This article covers technical details, affected versions, and mitigation.

Published:

CVE-2026-68578 Overview

CVE-2026-68578 is an authentication bypass vulnerability in ArcadeDB versions before 26.7.3. The flaw resides in the Model Context Protocol (MCP) HTTP transport, which fails to bind the authenticated principal to the request context. As a result, engine permission checks silently pass as no-ops for MCP-allowed users. Attackers with non-root MCP access can perform arbitrary database writes, execute Data Definition Language (DDL) statements, mutate schemas, and run arbitrary JavaScript code through the query tool. The weakness is classified as [CWE-306] Missing Authentication for Critical Function.

Critical Impact

Low-privileged MCP users gain unrestricted control over ArcadeDB, including arbitrary JavaScript execution via the query tool.

Affected Products

  • ArcadeDB versions prior to 26.7.3
  • ArcadeDB deployments exposing the MCP HTTP transport
  • Any ArcadeDB instance permitting non-root MCP-allowed users

Discovery Timeline

  • 2026-08-02 - CVE-2026-68578 published to the National Vulnerability Database (NVD)
  • 2026-08-03 - Last updated in NVD database

Technical Details for CVE-2026-68578

Vulnerability Analysis

ArcadeDB exposes an MCP HTTP transport that accepts tool invocations from authenticated clients. In vulnerable builds, the transport layer authenticates the incoming request but does not propagate the authenticated principal into the database engine's security context. Downstream engine permission checks therefore evaluate against a null or default principal and return success without enforcing role-based restrictions.

The consequence is a complete collapse of authorization inside the MCP execution path. A user who possesses only MCP access can invoke database operations that require elevated privileges. This includes destructive operations against production data, schema alterations, and code execution primitives exposed through the query tool.

The query tool accepts JavaScript payloads for scripted queries. Because permission checks act as no-ops, an attacker can execute arbitrary JavaScript inside the database server process. This transforms an authorization flaw into a code execution risk affecting confidentiality, integrity, and availability.

Root Cause

The root cause is a missing principal binding step in the MCP HTTP transport. Authentication succeeds at the transport boundary but the resulting identity is never attached to the engine session used for permission enforcement. The engine's authorization checks therefore have no subject to evaluate and default to permitting the operation.

Attack Vector

Exploitation requires network access to the MCP HTTP endpoint and valid low-privileged MCP credentials. An attacker authenticates as any non-root MCP-allowed user, then issues MCP tool calls such as query with DDL statements, write operations, or embedded JavaScript. The engine executes each request as if it originated from a fully privileged principal. See the GitHub Security Advisory and VulnCheck Advisory for further technical details.

Detection Methods for CVE-2026-68578

Indicators of Compromise

  • Unexpected CREATE, ALTER, or DROP statements executed by non-administrative accounts through the MCP endpoint.
  • MCP query tool invocations containing JavaScript payloads from users that should not possess script execution rights.
  • Schema mutations or bulk writes correlated with MCP HTTP transport request logs.

Detection Strategies

  • Audit ArcadeDB server logs for MCP tool calls originating from low-privileged users that perform DDL or schema changes.
  • Compare the authenticated principal recorded at the HTTP layer against the effective principal used by the engine; a mismatch indicates exploitation.
  • Alert on any use of the MCP query tool that submits JavaScript blocks, particularly from accounts not designated for administrative scripting.

Monitoring Recommendations

  • Forward ArcadeDB access and audit logs to a centralized analytics platform for correlation with identity data.
  • Baseline normal MCP tool usage per user and flag deviations such as first-time DDL or script invocations.
  • Monitor outbound network activity from the ArcadeDB host for signs of JavaScript-driven data exfiltration.

How to Mitigate CVE-2026-68578

Immediate Actions Required

  • Upgrade ArcadeDB to version 26.7.3 or later on every instance that exposes the MCP HTTP transport.
  • Restrict network access to the MCP endpoint to trusted management networks until patching is complete.
  • Rotate credentials for any MCP-allowed accounts and review recent MCP activity for signs of abuse.

Patch Information

ArcadeDB 26.7.3 addresses the vulnerability by binding the authenticated principal to the engine session used by the MCP HTTP transport. Refer to the GitHub Security Advisory GHSA-6x73-v3rc-f57c for the official fix and release notes.

Workarounds

  • Disable the MCP HTTP transport in ArcadeDB configuration if it is not required for operations.
  • Remove MCP allow-list entries for non-administrative accounts until the upgrade is deployed.
  • Enforce network-layer controls, such as firewall rules or reverse-proxy allow-lists, to block untrusted clients from reaching the MCP endpoint.
bash
# Configuration example: block external access to the MCP HTTP endpoint
# Adjust the port to match your ArcadeDB MCP transport configuration
iptables -A INPUT -p tcp --dport 2480 -s 10.0.0.0/8 -j ACCEPT
iptables -A INPUT -p tcp --dport 2480 -j DROP

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

Default Legacy - Prefooter | Experience the World’s Most Advanced Cybersecurity Platform

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.