Attack Chain

1. An attacker identifies a published Budibase application that uses a non-SQL datasource (MongoDB, CouchDB, Elasticsearch, DynamoDB-PartiQL, or REST-with-JSON-body) and has a PUBLIC role assigned to at least one query (e.g., GetUserByName).
2. The attacker, unauthenticated, sends an HTTP POST request to /api/v2/queries/:queryId for the identified public read query, including an x-budibase-app-id header.
3. The request body contains a JSON payload with the parameters field where a parameter (e.g., name) is injected with a crafted string like x",\"name\":{\"$exists\":true},\"$comment\":\"audit".
4. Budibase's enrichContext function substitutes this raw parameter value directly into the query's JSON body string without proper JSON-string escaping.
5. The subsequent JSON.parse operation on the malformed string results in a parsed filter object (e.g., collection.find()) where the attacker's injected $exists:true clause overrides the legitimate filter, effectively widening the scope to return all documents.
6. The Budibase server returns all documents from the collection, including sensitive fields not intended for public exposure (e.g., password_hash, secret), to the unauthenticated attacker.
7. If a PUBLIC write query (e.g., updateMany) is available, the attacker can similarly inject a payload to widen the filter scope, causing the updateMany operation to affect all documents in the collection, leading to widespread data modification or destruction.

Impact

Successful exploitation of CVE-2026-54350 allows an unauthenticated visitor to conduct broad data breaches and integrity compromises. This includes the anonymous reading of every document within any MongoDB, CouchDB, Elasticsearch, DynamoDB-PartiQL, or REST-with-JSON-body collection exposed via a PUBLIC query. Attackers can exfiltrate sensitive data such as password_hash, secret, api_token, and mfa_secret from columns not intended to be returned. Furthermore, where a PUBLIC update, delete, or aggregate query exists, attackers can anonymously modify or destroy every document in that collection, extending beyond the original builder's intended single-document scope. All of this can be achieved with a single HTTP request, bypassing session and CSRF protections.

Recommendation

• Patch CVE-2026-54350 immediately: Upgrade your Budibase server instances to version 3.39.12 or newer to remediate the vulnerability. Refer to the official Budibase release notes and patching guidance.
• Deploy the provided Sigma rule: Implement the Detect CVE-2026-54350 Exploitation Sigma rule in your SIEM to identify attempts at NoSQL operator injection against your Budibase applications.
• Monitor webserver logs: Specifically look for POST requests to /api/v2/queries/:queryId containing JSON body parameters with suspicious injection patterns (e.g., ",\\\"$exists\\\":true") in cs-uri-query or equivalent fields.
• Audit public queries: Review all PUBLIC role-assigned queries in your Budibase applications, especially those interacting with MongoDB, CouchDB, Elasticsearch, DynamoDB-PartiQL, or REST-with-JSON-body datasources, to ensure sensitive data is not inadvertently exposed or modifiable.