Postiz
Security release Postiz 2.25.3 closes critical file-read and token-forgery flaws
Registered users can exploit unrestricted provider methods to read backend files and crash the service (CVE-2026-115931), traverse post-media paths to read worker files (CVE-2026-115932), or obtain signed tokens for account and organization takeover (CVE-2026-115933). Exposed files can include JWT signing secrets and provider credentials.[2][3][4]
Fix. Upgrade to 2.25.3 to cover all three flaws; 2.25.2 fixes CVE-2026-115931 but not the other two.[1][2][3][4]
Versions, remediation & sources
- Affected systems & conditions
- CVE-2026-115931 affects 2.23.0 through versions before 2.25.2; CVE-2026-115932 affects 2.23.0 through versions before 2.25.3; CVE-2026-115933 affects 2.15.0 through versions before 2.25.3. An ordinary account is required; self-registration is open by default.[2][3][4]
- Remediation
- Upgrade to 2.25.3 to cover all three flaws; 2.25.2 fixes CVE-2026-115931 but not the other two. Setting DISABLE_REGISTRATION=true only reduces exposure to existing account holders. Leaving Discord credentials unset removes the main media-exfiltration route, not the alternate routes. If browser-extension channels are unused, blocking POST /integrations/extension-refresh removes the main token-signing route, not the invitation route. These workarounds do not replace the upgrade.[1][2][3][4]
- Evidence at review time
- The CNA records describe researcher proof-of-concept demonstrations on unmodified builds, including file extraction, backend termination and cross-organization takeover. They do not state malicious exploitation in the wild.[2][3][4]
- Related context
- The main media-exfiltration route requires a configured Discord provider; its credentials are empty in the shipped compose configuration. A validation route instead needs Dribbble, TikTok or Pinterest. The main token-forgery route requires outbound access to api2.skool.com. With Stripe billing configured, the attacker's plan must allow a channel and the target organization must not be on the STANDARD tier; session and password-reset token abuse is not subject to these limits. CVE-2026-115931 was published October 23 at 22:02 UTC; the other two records and the 2.25.3 release followed October 24.[2][3][4][1]