Description

Flowise before 3.1.4 fails to sanitize the overrideConfig.sessionId parameter before using it in MongoDB queries within the MongoDBMemory node. Unauthenticated attackers can submit MongoDB operator objects through the prediction API to read chat history records belonging to other users from the shared collection.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "7.5",
  "pubDate": "2026-09-15T16:17:45.170Z",
  "pubdate": "2026-09-15T16:17:45.170Z",
  "executiveSummary": "The Flowise application, in versions prior to 3.1.4, contains a critical NoSQL injection vulnerability within the MongoDBMemory node.\nThis vulnerability stems from improper input sanitization of the 'overrideConfig.sessionId' parameter, which is subsequently processed by MongoDB query operations.\nBy injecting malicious MongoDB operator objects into the prediction API, unauthenticated attackers can manipulate backend queries to bypass access controls.\nThe primary impact of this flaw is unauthorized data exfiltration, specifically allowing attackers to retrieve chat history records belonging to other users from shared MongoDB collections.\nThe vulnerability represents a high-risk security flaw as it does not require prior authentication or elevated privileges, effectively exposing sensitive historical user data to external threat actors.\nOrganizations utilizing Flowise must prioritize upgrading to version 3.1.4 or later to ensure that user-supplied input is correctly validated and sanitized before being passed to database driver functions.",
  "technicalDetails": "The vulnerability resides in the MongoDBMemory node implementation within Flowise. The application fails to sanitize or validate the 'overrideConfig.sessionId' parameter before passing it as a criterion for database queries.\nBecause the application uses this parameter directly in its interaction with the MongoDB database, an attacker can leverage MongoDB's query operator syntax to influence query logic. Specifically, by submitting a JSON object containing query operators (such as $ne, $regex, or $gt) instead of a simple string for the sessionId, the attacker can force the database to return records outside of their intended scope.\nThe attack flow initiates via the Flowise prediction API. An attacker crafts a malicious request payload where the 'sessionId' field within the 'overrideConfig' object is replaced with a MongoDB operator payload. For instance, providing a payload like {\"sessionId\": {\"$ne\": null}} instructs the MongoDB query engine to ignore typical session constraints and potentially return any record that meets the modified logic.\nThe vulnerable component is the database interaction layer within the MongoDBMemory node. By manipulating these queries, an attacker circumvents the application's intended data isolation, which is designed to partition chat history based on specific session identifiers.\nThis vulnerability is particularly severe because it is exploitable without requiring authentication, meaning any network-reachable instance of Flowise is susceptible to unauthorized data retrieval. The lack of input sanitization acts as a classic NoSQL injection vector, where the query structure is dynamically altered by user-supplied input that is improperly interpreted as code or database instructions rather than literal data.\nPost-exploitation, the attacker can successfully exfiltrate chat transcripts and other historical session data stored within the MongoDB instance. This leads to a total compromise of confidentiality for user communications stored in the affected database collections."
}