Skip to content

Security: pavan-dangeti/ScamShield-AI

SECURITY.md

Security policy

Reporting a vulnerability

Please report security issues privately through GitHub private vulnerability reporting, not in a public issue. Include the affected endpoint or file, steps to reproduce, and the impact you expect. You can expect an acknowledgement within a week.

Scope

In scope: the API in src/, the web UI in static/, the Twilio webhook, and the data pipeline's handling of personal data.

Out of scope: the free-tier demo's availability, and model mistakes on individual messages (please open an issue for those; misclassifications are useful data).

How the service protects itself and its users

  • Twilio webhook requests must carry a valid signature; validation is on by default.
  • CORS is restricted to configured origins, /api/analyze is rate limited, and input length is capped.
  • Errors return a request ID, never internal detail.
  • Phone numbers, email addresses, UPI handles and long identifiers are removed before a message is stored, and message text is never written to logs.
  • No credentials are committed; configuration comes from the environment (see .env.example).

There aren't any published security advisories