A breach is rarely a single step. An attacker lands on one low-value machine and then moves sideways — host to host, credential to credential — until they reach what they actually came for: domain controllers, data, and admin accounts. This page maps that journey onto the MITRE ATT&CK framework: the real techniques, the tools that execute them, how each is detected, and the controls that break the chain.
Frameworks exist so defenders and attackers describe the same thing the same way. The Cyber Kill Chain (Lockheed Martin) gives the high-level arc; MITRE ATT&CK breaks it into concrete, observed techniques, each with an ID like T1021. Lateral movement is tactic TA0008 — but it only matters because of what comes before and after it.
| Phase | ATT&CK tactic | What happens |
|---|---|---|
| Recon | TA0043 Reconnaissance | Map the target — domains, hosts, people, exposed services |
| Initial access | TA0001 Initial Access | Get the first foothold (phish, exploit, valid creds) |
| Execution & persistence | TA0002 / TA0003 | Run code; survive reboots and logouts |
| Credential access | TA0006 Credential Access | Steal hashes, tickets, tokens from the foothold |
| Lateral movement | TA0008 | Use those credentials to reach the next host |
| Privilege escalation | TA0004 | Climb from user → local admin → domain admin |
| Collection & exfiltration | TA0009 / TA0010 | Gather and remove the target data |
Lateral movement is credential reuse plus legitimate protocols. Attackers rarely drop new exploits to move internally — they log in. The traffic looks like normal admin activity, which is exactly why it's hard to spot and why segmentation and least privilege matter more than any single detection.
A flat network — where every host can talk to every other — means one foothold reaches everything. Zoning divides the network into trust tiers with controlled crossing points, so a compromise in one zone doesn't automatically become a compromise of all of them.
Internet-facing services (web, mail, VPN) live here, isolated from the internal network. If a public server is popped, the attacker is still boxed out of the core.
Workstations and everyday systems. The most common landing zone for phishing, and the launch pad for moving deeper.
Application, database and file servers. Should only accept the specific traffic it needs — not general access from every workstation.
Domain controllers, credential stores, backups, admin infrastructure. Compromise here = game over. Treated as its own hard boundary.
Lateral movement needs a starting point. The overwhelming majority of real intrusions begin with one of three things, none of which is a Hollywood "hack":
Several of the worst breaches in history started at step one here. EternalBlue (CVE-2017-0144, MS17-010) — an SMB flaw — gave WannaCry and NotPetya both their entry and their worm-like spread in 2017. A single unpatched edge service can be the whole story.
Once an attacker holds credentials from the foothold, these are the techniques they use to reach the next machine. Each row is a mapped ATT&CK technique — learn the mechanism, not just the name.
| Technique / ID | How it works | Tools | Detection signal |
|---|---|---|---|
| Pass-the-Hash T1550.002 | Authenticate with the NTLM hash directly — the plaintext password is never needed | mimikatz, impacket, NetExec | NTLM logons (Event 4624 type 3) from unusual source hosts |
| Pass-the-Ticket T1550.003 | Inject a stolen/forged Kerberos ticket into a session to impersonate a user | mimikatz, Rubeus | Tickets with anomalous lifetimes; TGT/TGS mismatches |
| Kerberoasting T1558.003 | Request service tickets (TGS) for SPN accounts, crack them offline for the service password | Rubeus, impacket-GetUserSPNs, hashcat -m 13100 | Bursts of TGS requests (Event 4769) with RC4 encryption |
| AS-REP Roasting T1558.004 | Target accounts with Kerberos pre-auth disabled; crack the returned AS-REP offline | Rubeus, impacket-GetNPUsers, hashcat -m 18200 | AS-REQ for accounts that shouldn't be queried |
| Credential reuse T1078 | A shared local-admin password or reused user password opens the next box outright | NetExec, spray scripts | Same account authenticating across many hosts quickly |
These abuse legitimate remote-administration protocols — the same ones sysadmins use daily — which is what makes them blend in.
| Sub-technique / ID | Protocol / port | Typical use |
|---|---|---|
| Remote Desktop Protocol — T1021.001 | RDP · 3389 | Interactive graphical session to another host |
| SMB / Windows Admin Shares — T1021.002 | SMB · 445 | PsExec-style remote execution via ADMIN$ / C$ |
| Distributed COM — T1021.003 | DCOM · 135 | Execute via COM objects (e.g. MMC20) |
| SSH — T1021.004 | SSH · 22 | The Linux/Unix equivalent of RDP movement |
| Windows Remote Management — T1021.006 | WinRM · 5985/5986 | PowerShell Remoting; evil-winrm with creds/hash |
# pass-the-hash to a SYSTEM shell over SMB (T1021.002 + T1550.002)
impacket-psexec -hashes :<NThash> DOMAIN/admin@10.0.0.25
# dump every secret from a domain controller (needs DA / DCSync rights)
impacket-secretsdump DOMAIN/admin@dc01 -just-dc
# spray one hash across a whole subnet to find where it is admin
nxc smb 10.0.0.0/24 -u admin -H <NThash> --local-auth
Every technique above produces a logon event. The detection isn't a magic signature — it's behavioural: a service account logging on interactively, one account touching 40 hosts in a minute, NTLM where Kerberos is expected, RC4 tickets in an AES environment. This is what "unusual east-west traffic" on a SIEM dashboard actually means.
In most enterprises, Active Directory is the single source of truth for who can access what. Compromising it compromises everything downstream — so attackers treat reaching Domain Admin (or the krbtgt account) as the objective. A typical path:
krbtgt.With the krbtgt hash an attacker forges a Golden Ticket (T1558.001) — a self-signed TGT granting access as anyone, to anything, that survives password resets. A Silver Ticket (T1558.002) forges a service ticket for one specific service. This is why recovering from a true DA compromise often means rotating krbtgt twice — and sometimes rebuilding the forest.
Movement and escalation interleave. Each new host may offer a local privilege-escalation route that yields fresh, higher-value credentials to move again:
sudo / SUID binaries (Linux) — GTFOBins catalogues which allow a shell escape to root.SeImpersonate to jump to SYSTEM.Tools like WinPEAS / LinPEAS automate the "what can I abuse here" check on a new host — the same checks a defender should run to find and close the gaps first.
You can't prevent every foothold — so the goal is to make sure one foothold stays one foothold. These controls directly map to the techniques above.
| Control | What it stops |
|---|---|
| Network segmentation & VLANs | Removes the flat-network assumption; a DMZ foothold can't reach the server zone |
| Least privilege + LAPS | Unique local-admin passwords per host kill credential-reuse spraying |
| Tiered administration | Tier-0 (DC) admins never log on to workstations, so their creds never land where they can be stolen |
| MFA everywhere | A stolen password alone no longer grants access — especially on VPN, RDP and privileged logins |
| Disable/limit NTLM; enforce Kerberos + AES | Removes pass-the-hash surface and RC4 roasting weakness |
| EDR + SIEM on east-west traffic | Detects the behavioural anomalies these techniques generate |
| Patch exposed & internal services | Closes the worm routes (EternalBlue, BlueKeep CVE-2019-0708, PrintNightmare, Zerologon CVE-2020-1472) |
| Credential Guard & LSASS protection | Stops the memory-dump step that feeds everything downstream |
SEGMENT · LEAST PRIVILEGE · MFA · DETECT · PATCH