Server monitoring · Security signs

Signs that a server is exposed or has been broken into

Most break-ins are not noticed for weeks. The same read-only agent that shows CPU and memory also looks for the cheap, well-known signs, and tells you once when one appears.

What it looks for

  • A database or admin service listening on every address: Redis, PostgreSQL, MySQL, MongoDB, Elasticsearch, the Docker API and others. When a Docker container publishes the port, the alert names the container.
  • A port that was not open before and stays open.
  • A new account that can log in, or an extra account with root's id.
  • Many SSH logins in progress at once, which usually means passwords are being guessed.
  • Many outgoing mail connections from a server that is not a mail server: the usual sign of a spam script or mail worm.
  • Many outgoing SSH or Telnet connections, or connections to crypto-mining ports.
  • A program named like a part of the Linux kernel, or running from /tmp or a hidden folder, even when it uses almost no CPU.
  • /etc/ld.so.preload being set, which some rootkits use to hide.

Also shown on the server's page

  • Whether SSH allows root login and password login.
  • Every port open on every address.
  • Services that failed, programs killed for lack of memory, and whether a restart is needed to finish security updates.

How it avoids false alarms

  • The first day is spent learning. Ports and accounts that are already there count as normal; only new ones are reported.
  • A new port must stay open for 15 minutes. SSH guessing and outgoing mail must last a few minutes.
  • One message when a sign appears. The alert is off until you switch it on; the page shows the findings either way.

What it is not

These are signs, not proof, and each message says what to check. It does not scan files for malware, read your logs or block anything, and it cannot see past a firewall in front of the server. It is not an antivirus, a firewall or an intrusion detection system. It needs no extra permission: everything is read from files any user on the server can read, and only names, ports and counts are sent.

Questions

Frequently asked questions

The honest answers, before you sign up.

1

Does it tell me who is attacking?

No. No addresses are collected. It tells you that something is happening and what to look at.

2

My database port is open but a firewall blocks it. Will it still alert?

Yes. The agent sees the server from inside and cannot know about a firewall in front of it. The message says so, and you can leave the alert switched off.

3

Does it need root?

No. The agent runs as its own user without sudo.