Skip to content

Image: securitylabs.dd-static.net · rights & removal

Executive Summary

A critical use-after-free vulnerability, CVE-2026-118383, exists in MongoDB Server when a newly elected primary completes its catch-up process, affecting replica set members used by shards and config servers. The vulnerability allows for remote unauthenticated exploitation, although the specific window of opportunity is not fully known. Reproduction steps involved setting up conditions where an attacker could plant stale "waiters" via unauthenticated ping requests before a failover occurs, leading to a segmentation fault in the server process during catch-up cleanup.
The advisory provides specific fixed release versions for various MongoDB branches, such as 8.0.37 for version 8.0 and 8.3.16 for version 8.3. This vulnerability is confined to replica set members; standalone mongod instances are unaffected. The potential impact includes the server terminating unexpectedly or executing arbitrary code under certain conditions. Mitigation involves upgrading to fixed releases and restricting network access, although the latter does not fully resolve the underlying memory safety issue.

Facts Only

* CVE-2026-118383 is a critical use-after-free in MongoDB Server's primary catch-up process.
* The vulnerability affects replica set members, including those used by shards and config servers in sharded clusters.
* Standalone mongod instances are unaffected.
* An unauthenticated client can induce a crash via two ping requests before login, as vulnerable servers accept ping before login.
* A crash (segmentation fault) was reproduced on MongoDB Server 8.0.34.
* Exploitation occurs when a newly elected primary finishes catch-up and performs cleanup.
* Fix releases include 8.0.37, 8.3.16, 9.0.7, and others.
* Exploitation involves planting stale "waiters" via unauthenticated requests that persist until the server restarts or steps down.
* Log analysis should look for commands with unusual operation-time values (e.g., `readConcern.afterOpTime`) correlated with catch-up events and fatal signals involving `CatchupState::abortinlock()`.

Full Take

The narrative surrounding CVE-2026-118383 highlights the inherent fragility of state transitions in distributed systems when operations are not strictly synchronized across all operational phases. The vulnerability is not merely a bug; it represents a failure in robust memory management during concurrent system events like replication catch-up and leader election, allowing an attacker to induce a controlled crash by exploiting cleanup logic based on preemptively placed, unexpired state markers.
The mechanism described—planting waiters that persist through failovers until the final cleanup phase—suggests a systemic vulnerability in how MongoDB manages transient states during high-stakes operations. The fact that patching requires upgrading across multiple specific branches underscores the challenge of maintaining security hygiene in complex, rapidly evolving distributed environments. Furthermore, the reproducible nature of the crash, demonstrated by an unauthenticated attacker leveraging low-level network interactions (ping) to manipulate internal state, shifts the focus from mere configuration errors to deep architectural flaws concerning trust boundaries and state persistence across failures.
The focus on detecting exploitation through log correlation, specifically linking operational metrics like `readConcern.afterOpTime` with catastrophic failure signals (`SIGSEGV`), indicates a necessary shift in defensive posture: moving beyond signature-based detection of known exploits toward monitoring the internal choreography of system events themselves. The implication is that resilience requires not only patching code but also implementing controls that guard against subtle, state-dependent manipulations during asynchronous recovery processes, forcing operators to question whether current monitoring tools are sufficiently sensitive to capture these deep temporal dependencies in production environments.
Bridge Questions: If the attacker cannot achieve arbitrary code execution through this path, what other forms of system instability or data integrity loss might be subtly introduced by exploiting similar race conditions across different consensus protocols? How can security observability tools be designed to inherently monitor and flag asynchronous state transitions rather than just endpoint outcomes? What are the long-term consequences for trust when recovery mechanisms themselves become vectors for attack based on timing and sequence?

From the original · Datadog Security Labs

Key points and observations - CVE-2026-118383 is a critical use-after-free in MongoDB Server's primary catch-up process, with a CVSS v4.0 score of 9.5. It affects replica set members, including those used by shards and config servers.
Read the full story at securitylabs.datadoghq.com

Sentinel unavailable

The automated check of the source article's wording did not complete, so no result is shown.

This looks only at the wording of the original source article, not at this page's AI-written sections. A small local AI model made this estimate. It has not been checked against known human and machine texts, so treat it as provisional. It cannot show who wrote an article.

CVE-2026-118383: Use | Huntaegis