MongoDB Server — Critical Pre-Auth Flaw Can Crash Lagging Servers on Failover (CVE-2026-118383)

Publication date: November 12, 2026
Category: Vulnerability / Database Security

Introduction

A critical security vulnerability affecting MongoDB Server allows unauthenticated threat actors to disrupt the operational stability of distributed database clusters by transmitting maliciously crafted requests. The flaw, identified and reported by MongoDB alongside independent findings from Datadog Security Labs researchers, manifests specifically during the catch-up process following a failover event in lagging replica members. The vulnerability has been tracked officially as CVE-2026-118383 and assigned a base severity score of 9.5 (CRITICAL) on the CVSS v4.0 scale.

What is CVE-2026-118383? (General Analysis)

The affected component is the replication subsystem within MongoDB Server, which is responsible for maintaining data consistency and synchronization across nodes in a replica set. In high-availability MongoDB architectures, a designated primary node processes write operations while secondary nodes replicate the data. When the primary node becomes unavailable, remaining members hold an election to elect a new leader.

The flaw represents a classic use-after-free memory management error, officially categorized under CWE-416. It occurs due to flaws in cleanup logic when a newly elected, lagging replica member completes its catch-up phase. Official NVD data confirms that this anomalous condition can be triggered through requests that do not require prior authentication, resulting in the abrupt termination of the database server process (mongod) and potentially opening a theoretical path to arbitrary code execution.

  • Official CVSS Vector (CVSS v4.0): CVSS:4.0/AV:N/AC:H/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H (Score: 9.5 - Critical).
  • Official CWE Classification: CWE-416 (Use After Free).
  • CISA KEV Status: Not listed in the active exploitation catalog at the time of publication.

How Does It Work? (Technical Analysis)

The exploitation and trigger mechanism requires a precise sequence of operational events within the database cluster:

  • Unauthenticated Request Planting (Waiters): An unauthenticated attacker sends ping commands incorporating a readConcern.afterOpTime parameter. These requests instruct the server to wait until replication reaches a specific point, creating and recording an internal pending notification known as a waiter.
  • Waiter Persistence: The planted requests do not expire or purge automatically; they persist in memory on the targeted node until it restarts, undergoes a rollback, or steps down from its primary role. Enabling standard authentication on vulnerable versions does not prevent these requests from being processed and stored.
  • Failover Activation and Double Cleanup: For the vulnerability to trigger a crash, the member that received the planted requests must win a subsequent election despite lagging behind other peers. Upon completion of the catch-up phase, the system’s cleanup routine erroneously executes twice on the same previously freed memory block.
  • Memory Corruption and Termination: This double-free condition of control structures triggers a segmentation fault or a fatal crash signal involving the CatchupState::abort_inlock() routine, causing the new primary node to crash before accepting any new writes and leaving the cluster temporarily without a functioning leader.

Affected Systems / Environments

The issue directly impacts replica set members, explicitly including shard and config server nodes within sharded clusters.

CVE Category (CWE) Impact CVSS Vector (summary)
CVE-2026-118383 CWE-416 (Use After Free) Denial of Service (Crash) / Potential Code Execution 9.5 (Critical) AV:N/AC:H/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H
  • Unaffected Deployments: Standalone mongod instances operating outside a replica set configuration are not affected.
  • Lifecycle Status: Software branches prior to the vendor’s patched releases are vulnerable. Notably, the 8.2 branch reached End of Life on July 31, 2026, and will not receive official security backports, despite proof-of-concept tests demonstrating ongoing vulnerability.

Mitigation and Detection

Remediation

  • Version Upgrades: Immediately apply the official security releases provided by MongoDB across all supported branches and on every active replica set member, including containerized deployments and nodes that were offline during routine maintenance windows. MongoDB Atlas clusters receive patches automatically.
  • Network Access Restriction (Temporary Mitigation): In environments where immediate patching is unfeasible, strictly restrict network access to the database to trusted, authorized systems only, thereby limiting exposure to unauthenticated request injection.

Detection

Defensive security teams should audit server structured logs for specific indicators identified during vulnerability reproduction:

Defensive Intelligence Note: The isolated occurrence of a primary election or a catch-up exit is standard behavior in distributed databases; however, the sequential chain of unauthenticated slow query commands followed by duplicate exits and fatal signals serves as a strong indicator of anomalous activity or exploitation attempts.

  • Log Indicators:
    • Slow query log entries (ID 51803) for unauthenticated ping commands carrying readConcern.afterOpTime that end with MaxTimeMSExpired.
    • Transitions where a member enters primary catch-up (ID 21359) and subsequently exits it twice (ID 21363) within a single state transition.
    • Fatal signals or backtrace messages involving CatchupState::abort_inlock() immediately following those exits.

Wrapping Up

The CVE-2026-118383 vulnerability highlights the inherent complexity of state and memory management within high-availability database replication protocols. A subtle flaw in post-failover cleanup logic can result in critical service disruption via pre-auth attack vectors. Rigorous patch management policies and strict network segmentation remain indispensable defenses for ensuring availability in modern enterprise data infrastructures.

References

  • MongoDB. (2026). Important: MongoDB Security Update Available - Action Required. Bulletin Link
  • Datadog Security Labs. (2026). CVE-2026-118383: MongoDB Server Use-After-Free Technical Analysis. Technical Analysis Link
  • The Hacker News. (2026). Critical Pre-Auth MongoDB Flaw (CVE-2026-118383) Can Crash Lagging Servers on Failover. Original Source
  • Common Vulnerabilities and Exposures. CVE-2026-118383 Record. CVE Record
  • MongoDB. Support Policy Lifecycles. Lifecycle Schedule
  • Googlesky. (2026). Proof-of-Concept Code for CVE-2026-118383. GitHub Repository