SECURITY
Security architecture you can inspect.
SafeHuman uses layered controls rather than claiming perfect detection.
Account security
- Passwords use scrypt hashing with a unique random salt.
- JWT access tokens are signed with a production secret.
- TOTP multi-factor authentication is supported.
- Password-reset tokens are random, hashed in storage and expire.
Application security
- Trusted-host and CORS controls are configurable.
- Public scanner/auth endpoints have application-level rate-limit guardrails in this build.
- Uploads have byte-size limits and images are parsed before analysis.
- Business API keys are hashed, revocable and quota-limited.
- Business webhooks are signed and their destination URLs are validated.
- Production Docker runs the app as a non-root user.
Detection limitations
SafeHuman can make mistakes. A low-risk result is not proof that a site, message or sender is safe, and a high-risk result is not proof of malicious intent. Use independent verification for important financial, identity or account decisions.
Known architectural limitation
The browser client currently stores the bearer token in localStorage. That makes prevention of cross-site scripting especially important. A future session-hardening release should evaluate Secure, HttpOnly, SameSite cookies without breaking API clients.
Responsible disclosure
If you identify a security issue, report it privately through the operator’s official CyberHELP/SafeHuman contact channel. Do not publish credentials, user data or exploitable details before remediation coordination.
