CVE-2026-35676

Summary

Summary The password reset API can be triggered without authentication and without any out-of-band confirmation step. If an attacker knows a valid username + email pair, they can call the reset endpoint directly. The application immediately generates a new password, writes it to the account, and only then sends the new password by email. This creates two issues at the same time: account enumeration through the response difference between valid and invalid pairs forced password reset of another user's account, which invalidates the old password immediately In my local reproduction, I confirmed both the response difference and the password change itself. Details The relevant code is in phpmyfaq/src/phpMyFAQ/Controller/Frontend/Api/UnauthorizedUserController.php. The route is exposed without authentication: The flow is straightforward: The core issue is that the password is changed immediately after a simple username and email match. There is no reset token, no confirmation link, no second step, and no requirement that the caller prove control of the mailbox before the password is replaced. That means the endpoint is not just a "forgot password email sender". It is an actual unauthenticated password change trigger. PoC This was reproduced against a local Docker deployment of the project. For a valid username and email pair: Response: For an invalid pair: Response: That already confirms enumeration. To verify that the password really changes, created a test account: username: user2 password: Oldpass123! email: [email protected] Before calling the endpoint, the password hash stored in faquserlogin was: Then send: The response was: After that, the stored hash changed to: Then try to log in with the old password: The application redirected back to the login page and reported that the password was incorrect. So this is not just a cosmetic issue in the API response. The old credential really becomes invalid. Impact The most realistic impact here is forced password reset and account disruption. An attacker who knows or can guess a valid username and email pair can: confirm whether the pair is valid force the target account's password to change cause the victim's old password to stop working immediately If the attacker does not control the victim's mailbox, this is usually a denial-of-service style account disruption rather than instant account takeover. That is the main reason I would keep the severity at Medium. Even so, this is still a real security issue. Password recovery should not allow an unauthenticated caller to change the account password directly. Remediation Recommend to change the password recovery flow to a token-based design. Do not change the password inside the unauthenticated endpoint. Generate a short-lived, single-use reset token and send only the reset link by email. Return the same generic response for both valid and invalid username/email pairs. Keep rate limiting in place, but do not rely on it as the main protection. Add regression tests that verify the password hash does not change until a valid reset token is presented.

Impact

Affected versions

Security releases

Remediation advice