Summary
PraisonAI Platform contains an insecure direct object reference where project-level CRUD and stats endpoints authorize only on workspace membership, but do not bind project lookups to the workspace context.
As a result, any user who is a member of one workspace can access, modify, or delete projects that belong to other workspaces, making cross-workspace data exposure and tampering possible when project IDs are known or guessable.
PraisonAI Platform contains an insecure direct object reference where project-level CRUD and stats endpoints authorize only on workspace membership, but do not bind project lookups to the workspace context.
As a result, any user who is a member of one workspace can access, modify, or delete projects that belong to other workspaces, making cross-workspace data exposure and tampering possible when project IDs are known or guessable.
Affected context
This vulnerability occurs in PraisonAI Platform's project management APIs within multi-workspace deployments.
It is relevant wherever project resources are identified by global or easily discoverable IDs and access checks rely solely on workspace membership without verifying that the requested project belongs to the caller's workspace.
The risk is higher in environments where project IDs can be enumerated, logged, shared, or inferred from UI elements, logs, or external integrations.
This vulnerability occurs in PraisonAI Platform's project management APIs within multi-workspace deployments.
It is relevant wherever project resources are identified by global or easily discoverable IDs and access checks rely solely on workspace membership without verifying that the requested project belongs to the caller's workspace.
The risk is higher in environments where project IDs can be enumerated, logged, shared, or inferred from UI elements, logs, or external integrations.
Exploit preconditions
- Attack surface: HTTP API endpoints for project CRUD and stats:
GET /workspaces/{workspace_id}/projects/{project_id}PATCH /workspaces/{workspace_id}/projects/{project_id}DELETE /workspaces/{workspace_id}/projects/{project_id}GET /workspaces/{workspace_id}/projects/{project_id}/stats- Required access: Valid account that is a member of any workspace in the system.
- Trigger point: Supplying a
project_idassociated with a different workspace while passing a workspace ID for which the attacker has membership. - Why exploitation is possible: Authorization is performed as
require_workspace_member(workspace_id)only; the backing service methods (get,update,delete,get_stats) are invoked with a globalproject_idand noworkspace_idconstraint, so cross-workspace access is not prevented.
- Attack surface: HTTP API endpoints for project CRUD and stats:
GET /workspaces/{workspace_id}/projects/{project_id}PATCH /workspaces/{workspace_id}/projects/{project_id}DELETE /workspaces/{workspace_id}/projects/{project_id}GET /workspaces/{workspace_id}/projects/{project_id}/stats- Required access: Valid account that is a member of any workspace in the system.
- Trigger point: Supplying a
project_idassociated with a different workspace while passing a workspace ID for which the attacker has membership. - Why exploitation is possible: Authorization is performed as
require_workspace_member(workspace_id)only; the backing service methods (get,update,delete,get_stats) are invoked with a globalproject_idand noworkspace_idconstraint, so cross-workspace access is not prevented.
Why this happens
The underlying bug is an insecure direct object reference (IDOR), mapped to CWE-639: Authorization Bypass Through User-Controlled Key.
The API trusts the caller-provided project_id as a global identifier and performs authorization solely on the surrounding workspace resource, without verifying that the targeted project is actually owned by the authorized workspace.
This mismatch between authorization scope (workspace membership) and object lookup (unscoped project ID) allows users to reference and operate on projects belonging to other workspaces, bypassing intended tenant isolation.
The underlying bug is an insecure direct object reference (IDOR), mapped to CWE-639: Authorization Bypass Through User-Controlled Key.
The API trusts the caller-provided project_id as a global identifier and performs authorization solely on the surrounding workspace resource, without verifying that the targeted project is actually owned by the authorized workspace.
This mismatch between authorization scope (workspace membership) and object lookup (unscoped project ID) allows users to reference and operate on projects belonging to other workspaces, bypassing intended tenant isolation.
Potential impact
- A user may read detailed information about projects in other workspaces.
- A user may update configuration or metadata of another workspace's projects.
- A user may delete projects owned by other teams or organizations.
- A user could retrieve project statistics for workspaces they do not belong to, leaking usage patterns or performance data.
- Cross-tenant data exposure and integrity violations can undermine trust in multi-tenant deployments.
- A user may read detailed information about projects in other workspaces.
- A user may update configuration or metadata of another workspace's projects.
- A user may delete projects owned by other teams or organizations.
- A user could retrieve project statistics for workspaces they do not belong to, leaking usage patterns or performance data.
- Cross-tenant data exposure and integrity violations can undermine trust in multi-tenant deployments.
General mitigation guidance
Effective mitigation focuses on aligning authorization with resource ownership and constraining lookups to the caller's scope.
Common defensive patterns include:
- Scoping data access by tenant or workspace: require that any project lookup includes both
workspace_idandproject_id, and verify ownership at the data layer. - Enforcing object-level authorization checks that confirm the current principal's rights over the specific project, not just its workspace.
- Designing identifiers to be opaque and non-enumerable to reduce trivial guessing or scraping.
- Centralizing access control logic so that resource services cannot be called with unconstrained global IDs.
- Applying least-privilege principles, ensuring users only receive permissions necessary for operations within their own workspace.
Effective mitigation focuses on aligning authorization with resource ownership and constraining lookups to the caller's scope.
Common defensive patterns include:
- Scoping data access by tenant or workspace: require that any project lookup includes both
workspace_idandproject_id, and verify ownership at the data layer. - Enforcing object-level authorization checks that confirm the current principal's rights over the specific project, not just its workspace.
- Designing identifiers to be opaque and non-enumerable to reduce trivial guessing or scraping.
- Centralizing access control logic so that resource services cannot be called with unconstrained global IDs.
- Applying least-privilege principles, ensuring users only receive permissions necessary for operations within their own workspace.
How teams detect issues like this
Detection typically involves reviewing both code and runtime behavior for authorization gaps around user-controlled identifiers.
Useful strategies include:
- Static analysis or manual code review to identify endpoints where authorization is performed on a container resource (e.g., workspace) but object lookups use unconstrained IDs without verifying ownership.
- Searching for service methods that take global IDs (such as
project_id) without a corresponding tenant or workspace parameter. - Adding automated tests that simulate cross-tenant requests, attempting to access resources from another workspace while authenticated as a different workspace member.
- Logging and monitoring access patterns to detect anomalies, such as a single user accessing large numbers of projects across multiple workspaces.
- Incorporating authorization policy engines or middleware that enforce consistent tenant scoping across all resource access paths.
Detection typically involves reviewing both code and runtime behavior for authorization gaps around user-controlled identifiers.
Useful strategies include:
- Static analysis or manual code review to identify endpoints where authorization is performed on a container resource (e.g., workspace) but object lookups use unconstrained IDs without verifying ownership.
- Searching for service methods that take global IDs (such as
project_id) without a corresponding tenant or workspace parameter. - Adding automated tests that simulate cross-tenant requests, attempting to access resources from another workspace while authenticated as a different workspace member.
- Logging and monitoring access patterns to detect anomalies, such as a single user accessing large numbers of projects across multiple workspaces.
- Incorporating authorization policy engines or middleware that enforce consistent tenant scoping across all resource access paths.

