Doogree
Defense · NDR

Catching Lateral Movement Before It Reaches the Crown Jewels

07/06/2026 · ~6 min read
NDRLateral MovementT1021Logon-DNA

Attackers almost never land directly on the crown jewels. They gain a foothold on one machine — an ordinary, unimportant user workstation — and from there they crawl sideways, machine to machine, until they reach the file server or the DC. Classic antivirus and EDR look at a single host; they miss exactly what happens between hosts. NDR looks at the traffic itself.

Why east-west is the blind spot

A typical attack is a chain: initial access (phishing, stolen credential) → lateral movement (east-west, machine to machine) → impact (ransomware encryption, data theft). A security tool that looks at one machine sees a process running and a port listening — but it doesn't see that this machine is suddenly speaking SMB to ten other machines it never spoke to before. That's the signature of lateral movement, and it's visible only when you look at the internal network as a system.

Why this is critical for a small business: lateral movement is the middle of the attack — the part between the foothold and the encryption. That's exactly the window in which you still have time to stop the ransomware before it touches your files.

How we catch lateral movement

Doogree historizes the internal peer connections of every machine — the IP, the port, and the process that initiated the connection. For each connection we classify the port to an ATT&CK technique: SMB → T1021.002, RDP → T1021.001, WinRM → T1021.006, SSH → T1021.004, VNC → T1021.005. After the learning window, a first-ever internal peer on an admin/remote-control port is a lateral-movement candidate — and it surfaces as a behavioral DNA anomaly, tagged with the precise technique.

Logon identity as DNA (servers)

The flip side of the same coin is who logs in, and from where. The server agent reports the human accounts and the successful logins from auth.log — the account, the source IP address, and whether it's a root/admin login. The cloud builds a baseline of the known accounts and the known logon sources. A new account, or an admin/root login from an IP that's never been seen, is a critical anomaly — an account-takeover candidate — mapped to T1078 (Valid Accounts) and T1098.

agentless-leaning. Collecting logon identity relies on a server-side collector, not a heavy agent on every machine. This is the agentless-first approach: the weight is in the cloud, the touch at the endpoint is minimal and non-disruptive.

It all folds into one attack story

These two capabilities don't live alone. They fold into the SIEM correlation engine as stages in the kill chain: the lateral movement is the lateral stage, and the logon-identity anomaly is the access stage. Instead of two isolated alerts, you get one incident that tells a story: someone logged in from a foreign source, then crawled sideways over WinRM — with automatic severity escalation.

Why this beats single-host EDR: EDR sees a process; NDR sees the traffic. Combining port-to-technique tagging with a baseline of internal peers and of logon sources gives you what most small businesses lack — visibility into the middle of the attack, at the point where you can still stop it. All on the same agent and the same collector that are already installed.

← Back to the blog · How it works →