libmongocrypt CVE-2026-81523

Exploitation requires: (1) the target application must have MongoDB Client-Side Field Level Encryption (CSFLE) configured with automatic encryption enabled - this is a non-default, explicitly configured feature; (2) the attacker must have local access and low privileges sufficient to influence the database or collection name argument passed to the libmongocrypt automatic-encryption context (AV:L, PR:L from the CVSS 4.0 vector); (3) specific attack prerequisites must be satisfied (AT:P), meaning the attacker cannot trivially reach the vulnerable code path without controlling an input at the application or driver API layer. Limiting factors: no network-based exploitation is possible (AV:L), unauthenticated remote attackers cannot trigger the flaw, and the vulnerability is entirely inert in MongoDB deployments not using CSFLE automatic encryption.

The risk signals are uniformly low. CVSS 4.0 scores 2.0, with AV:L (local attack vector - the attacker must control input within the application process or calling context), AT:P (attack prerequisites present - automatic encryption must be configured and in use), PR:L (low privileges required), and impacts limited to VC:L/VI:L with no availability impact. No KEV listing and no public exploit code exist at time of analysis. EPSS data was not provided. The vulnerability's scope is further constrained to deployments actively using MongoDB's CSFLE automatic encryption feature, which is not a default or universally deployed capability. The most realistic concern is an insider or compromised application component that can supply crafted namespace strings to the encryption library, resulting in schema confusion rather than direct data exfiltration or integrity compromise. This does not require network access and cannot be triggered remotely against a default MongoDB deployment.

A low-privileged caller - such as a compromised microservice or a malicious internal application component - supplies a crafted database name containing '.' or a NUL byte to the libmongocrypt automatic-encryption context setup API. The unsanitized input causes the library to resolve the wrong namespace, selecting an unintended encryption schema for the operation. The caller can then read or write data fields using an incorrect encryption policy, enabling limited schema-level information disclosure or modification within the application's data handling context. No public exploit code exists at time of analysis.