CVE-2026-96299: 
MongoDB 脆弱性分析と緩和策

CVE-2026-96299 is a denial-of-service vulnerability in MongoDB's mongos sharded cluster router component, where an authenticated user with readWrite privileges can crash the mongos process by crafting a specially formed aggregation pipeline. The vulnerability was published on October 13, 2026, and affects MongoDB Server versions 8.3.0 through 8.3.12 and 9.0.0 through 9.0.3. It carries a CVSS v3.1 base score of 6.5 (Medium) and a CVSS v4.0 base score of 7.1 (High) (GitHub Advisory, Feedly).

The root cause involves two weaknesses: CWE-617 (Reachable Assertion) and CWE-770 (Allocation of Resources Without Limits or Throttling). An internal aggregation stage in mongos lacks proper authorization checks despite being intended exclusively for internal use; an attacker can invoke this stage in a user-supplied aggregation pipeline, forcing a subsequent stage to execute on the router under conditions it was not designed to handle. This triggers an unhandled fatal assertion error, causing the mongos process to immediately abort. A secondary description from ENISA/EUVD notes that internal schema validation operators accepted in user-supplied query filters, when nested and applied to array-containing documents, can also cause unbounded memory consumption until the process terminates (GitHub Advisory, MongoDB Jira).

Successful exploitation results in an immediate, complete denial of service for all clients routed through the affected mongos instance, as the process aborts without recovery. There is no confidentiality or integrity impact — data is not exposed or modified. Because mongos acts as the entry point for all sharded cluster traffic, a single authenticated attacker can disrupt database availability for all downstream applications relying on that router (GitHub Advisory, Feedly).

No public proof-of-concept exploit code or in-the-wild exploitation has been reported as of the disclosure date. The EPSS score is 0.0, and the vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog. Exploitation requires valid credentials with at least readWrite privileges, which limits the attacker pool but does not eliminate risk in multi-tenant or shared environments (GitHub Advisory, Feedly).

1. Obtain credentials: Acquire valid MongoDB credentials with at least readWrite privileges on a database hosted on a sharded cluster (e.g., via a compromised application account or a free-tier/trial account).
2. Connect to mongos: Establish a connection to the target mongos router endpoint using a MongoDB client (e.g., mongosh, PyMongo, or the MongoDB driver of choice).
3. Craft malicious aggregation pipeline: Construct an aggregation pipeline that invokes an internal-only aggregation stage (one lacking proper authorization enforcement) followed by a stage that is incompatible with execution on the router.
4. Execute the pipeline: Run the crafted aggregation query against a collection on the sharded cluster. The mongos process evaluates the pipeline, hits the unauthorized internal stage, and triggers an unhandled fatal assertion.
5. Achieve denial of service: The mongos process immediately aborts, severing all client connections routed through it and causing a denial of service for all dependent applications (GitHub Advisory, MongoDB Jira).

• Logs: Sudden mongos process termination entries in MongoDB logs (e.g., CONTROL severity messages indicating a fatal assertion or abort signal); log lines referencing assertion failures in aggregation pipeline execution.
• Process: Unexpected exit/restart of the mongos process without a corresponding system-level cause (OOM killer, hardware fault, etc.); monitoring alerts for mongos process crashes.
• Network: Abrupt disconnection of all client sessions routed through a specific mongos instance; connection timeout errors from application servers targeting the mongos endpoint.
• Application Logs: Downstream application errors indicating loss of database connectivity coinciding with mongos restart events; repeated aggregation query attempts from a single authenticated user prior to the crash (GitHub Advisory).

MongoDB has released patched versions addressing this vulnerability: 8.3.13 and 9.0.4. Users running MongoDB Server 8.3.0–8.3.12 or 9.0.0–9.0.3 in sharded cluster configurations should upgrade to the respective fixed versions immediately. As a temporary measure, restricting readWrite privileges to only fully trusted users and monitoring for unusual aggregation pipeline activity can reduce exposure until patching is feasible (GitHub Advisory, MongoDB Jira).