CVE-2026-46478: Flowise: DatasetRow create+update mass-assignment allows cross-workspace row takeover

Synopsis

A mass-assignment vulnerability in Flowise, the drag-and-drop LLM flow builder, allows an authenticated attacker to create or update DatasetRow records across workspace boundaries. The flaw is reachable over the network and requires only a low-privilege account, with no special conditions beyond an existing authenticated session. Successful exploitation gives the attacker full read, write, and availability control over targeted rows in other workspaces. A patched-image rebuild at version 3.1.2 is available on HarborGuard for environments running an affected version.

HarborGuard Coverage

Exploit Conditions

• Network reachabilityRequired

  The vulnerable API endpoint is exposed over the network, so an attacker must be able to reach the Flowise service directly.

• AuthenticationRequired

  A low-privilege account is sufficient; no admin credentials are needed to submit malicious mass-assignment payloads.

• Victim interactionNot required

  The attacker acts entirely through their own authenticated API requests and does not need any other user to take any action.

• Attack complexityDetail

  Base exploit logic is straightforward, though the CVSS vector notes an attack requirement (AT:P), meaning specific deployment or configuration conditions must be present for the cross-workspace takeover to succeed.

Blast Radius

• Reads DatasetRow contents belonging to other workspaces, potentially exposing training data, prompt datasets, or sensitive LLM pipeline inputs.
• Overwrites or tampers with DatasetRow records in other workspaces, corrupting datasets used to drive LLM flows.
• Disrupts the availability of targeted DatasetRows, breaking downstream LLM pipeline runs that depend on those records.
• Cross-workspace boundary violation allows lateral movement across tenant data within the same Flowise deployment.