Speakr, a self‑hosted audio transcription service, contains a flaw in its token authentication and CSRF exemption mechanism. The application treats any supplied token, including those received in URL query parameters, as valid without verification. An attacker can issue a crafted request that calls the csrf_exempt_for_api_tokens hook with a forged token, causing the framework to permanently exempt the targeted view from CSRF checks for all users. This allows a subsequent state‑changing request—such as a POST to /account—to be performed without the original CSRF token, effectively turning a CSRF guard into a no‑op. Because the exemption applies globally to the view, an attacker can alter profile data, transcription settings, and even grant themselves administrative rights through routes like admin_toggle_admin or change_password. The vulnerability is divided into two weaknesses: improper authentication of API tokens and CSRF token bypass.

Speakr 0.8.x versions older than 0.8.21‑alpha, distributed by murtaza‑nasir, are impacted. All builds that use the buggy src/app.py and src/utils/token_auth.py modules prior to the patch are vulnerable. Users running custom forks that have not addressed the issue remain at risk.

The flaw has a CVSS score of 7.1, indicating high severity. The EPSS score, less than 1 %, suggests low expected exploitation probability in the short term, and the vulnerability is not listed in the CISA KEV catalog. However, the attack path requires a network‑reachable attacker and the ability to send a query parameter with a token that the application will accept. The attacker must also control a sibling subdomain to satisfy documented cookie conditions. Once the CSRF exemption is installed, the attacker can execute authorized requests without client‑side confirmation, permitting unauthorized modifications of user data and privileges.