শেয়ার

Security checklist

What the platform does by itself, and what the person running it has to do.

Built in

  • SSRF guard. Every user-given target (monitor URLs and hosts, webhook URLs, integration hosts, DNS resolvers) is resolved and validated before anything touches the network: private, loopback, link-local, multicast and reserved ranges, localhost, .local, .internal and the cloud-metadata address are refused; file:// and other schemes, and credentials inside URLs, are rejected. HTTP checks pin the validated IP and follow redirects by hand (max 5), validating each hop again. Private targets can be allowed per installation only by the administrator.
  • Encryption at rest. Channel configurations (bot tokens, webhook URLs, phone numbers), HTTP auth passwords, bearer tokens and API keys in monitor configs, integration credentials, SMTP passwords and payment gateway secrets are stored with AES-256-GCM under AIM_SECRET_KEY.
  • Tokens are hashed. API tokens, probe tokens, agent tokens, registration tokens, password-reset tokens, remember-me and trusted-browser tokens, 2FA backup codes and webhook signing secrets are stored as SHA-256 hashes (secrets encrypted); only the prefix of an API token is shown afterwards.
  • Passwords use bcrypt (password_hash), with rehashing when PHP's default changes, 8 failed attempts lock the account for 15 minutes.
  • Rate limits stored in the database (so they hold across workers and hosts): sign-in 10 per 10 minutes per IP, registration 5 per hour, password reset 5 per 15 minutes, 2FA 8 per 10 minutes, heartbeat 600 per minute per IP, status-page subscriptions 5 per hour, probe and agent registration 10–20 per 10 minutes, and the API per token per minute.
  • CSRF tokens on every form and AJAX call; same-site, HTTP-only cookies; Secure on HTTPS.
  • Sessions are tracked in a table with device and IP: "sign out everywhere", automatic sign-out of sessions revoked elsewhere, and a new-device e-mail on sign-in from an unknown address.
  • Two-factor authentication: TOTP app or e-mail code, 8 backup codes, 30-day trusted browsers.
  • Audit log of every account change and admin action with IP and details.
  • Agents never execute commands; the only configuration they read is the interval, the service names to check and the paused flag. Probes run only the check types the application ships, get monitor definitions over authenticated HTTPS and verify the signature of self-update packages.
  • Errors are logged, never shown, unless AIM_DEBUG is on.
  • Output is escaped; user Markdown and HTML are not rendered raw.

What the administrator must do

  • HTTPS only. Put the site behind a valid certificate and redirect HTTP to HTTPS. Tokens travel in headers.
  • A strong secret. The installer generates a 64-hex-character AIM_SECRET_KEY; never replace it with something short, never commit it, and keep it in your backups.
  • Keep config.php out of git and out of the web. The web server must deny includes/, storage/, cron/, probe/, agent/, lang/ and installed.lock (the shipped .htaccess and the nginx example do). config.php mode 640, owned by the web user.
  • File permissions. The application files read-only for the web user except storage/, uploads/ and includes/ (only while installing; set it back to read-only afterwards). Delete install.php after installing.
  • Cron key. The web-cron URL carries the key shown on Admin → System health (rotate it there if it leaks); treat it like a password and prefer a shell cron.
  • AIM_DEBUG false on a live site.
  • Private targets off unless you need to monitor a LAN, and then only for the organisations you trust (the guard still refuses the metadata address).
  • Trust proxy on only when a reverse proxy or Cloudflare is really in front; otherwise visitors could forge their IP.
  • Mail from a real mailbox (SMTP with SPF/DKIM) so verification, reset and alert e-mails are delivered and not spoofable.
  • Admin accounts with 2FA on, few of them, reviewed in Admin → Users.
  • Backups of the database, config.php, uploads/ and storage/, tested by restoring once (Backup & recovery).
  • Updates: apply releases promptly (db_update.php after each), keep PHP and MariaDB patched; probes with self-update on follow the application.
  • Agents: install them from your own URL (the one-liner), never from a copy you did not read; the token file must stay root-only.
  • Review the audit log and the notification log now and then; failed logins and admin actions are there.
এই পেজটি কি সহায়ক ছিল?

একটি ক্লিক আমাদের এই ডকস উন্নত করতে সাহায্য করে।

এই পেজ শেয়ার করুনসাধারণ লিংক - কোনো ট্র্যাকার নেই, ক্লিক না করা পর্যন্ত কিছু লোড হয় না।