Skip to content

Security: studio-prisma/osticket-plugin-feedback

Security

SECURITY.md

Security Policy

Supported Versions

Only the latest minor release receives security fixes.

Version Supported
0.1.x yes

Reporting a Vulnerability

Use GitHub Security Advisories. Reports there are private by default. Don't open a public issue for a suspected vulnerability.

Include:

  • plugin version, osTicket version, the settings that matter (mode, scale, on a bad rating)
  • what an attacker gains, not only what misbehaves
  • reproduction steps, ideally against a clean install (dev/up.sh builds one)

Expected response: an acknowledgement within 7 days, an assessment within 30. This is a spare-time project, so that's the honest number, not a target.

In Scope

  • A rating stored without a valid link: a forged, altered, expired or guessed token that the click page accepts.
  • A link that rates another ticket than the one it was sent for.
  • A rating stored by a plain page view (GET, HEAD, a prefetch), i.e. without the form being submitted in a session.
  • A POST that stores anything without osTicket's CSRF token for that session.
  • Ticket data on the click page beyond the ticket number: subject, thread, names, addresses, status.
  • Markup from a comment rendered on the click page, in a thread note, in the agent mail, in the assignment alert or on the evaluation page.
  • The token key or SECRET_SALT in a response, a log line, an error message or the release package; a link that works while SECRET_SALT is the sample value.
  • The IP address stored in clear text.
  • Switching the question off or on for a ticket (scp/ajax.php/feedback/ticket/<id>/decline|allow) without an agent session, without osTicket's CSRF token, or for a ticket the agent can't access.
  • feedback_ask widening the rules: a request that makes the plugin ask a ticket the rules exclude (AI agent, department, status, domain, already asked).
  • A redirect to an address outside the help desk or a link built from the request's Host header.
  • Bypassing the rate limit with the same IP, or growing the plugin's tables without bound from outside.
  • Asking twice for one ticket, asking for a ticket an AI agent closed, a sub-ticket or an excluded department, help topic or domain.
  • Reopening or assigning more than once for one ticket because of repeated or parallel clicks.
  • Access to the evaluation page without an agent login, or its data for agents who aren't admins.
  • SQL injection through any parameter of the click page or the settings.

Out of Scope

These are documented design decisions, not findings:

  • Anyone who has the mail can rate. The link is the authorisation, like a password-reset link. A forwarded mail lets the recipient rate and change the rating while it's valid.
  • Agents with the "with the mail" note can rate. That setting copies the sent mail, links included, into an internal note. Every agent who can open the ticket can follow them. The default note has no links.
  • A link scanner that runs JavaScript submits the rating like a person would. The confirmation step stops scanners that only fetch the page; one that executes the page in a sandbox is indistinguishable from a click.
  • Rate limit per IP counts all customers behind one NAT together (60 requests per 10 minutes, about 20 ratings).
  • No revocation of single links other than switching the plugin off (all links) or deleting the row in <prefix>feedback_request.
  • Emoji removed from thread notes and angle brackets shown as ‹ ›: limits of osTicket's sanitizer and its 3-byte utf8 database connection, see the README.
  • A customer who replies to the feedback mail reopens the ticket the way any reply to a ticket mail does. That's osTicket's behaviour.
  • osTicket core issues, other plugins, server and mail configuration.

There aren't any published security advisories