Quick answer
NTLM relay is an attack where an attacker passes a Windows login they intercepted on to another server, and is logged in there as the victim. Coercion techniques such as PetitPotam, PrinterBug and DFSCoerce force a server, often a domain controller, to log in to the attacker, which turns relay into a path to domain compromise.
On this page
NTLM is Windows' older login protocol. It proves who you are with a challenge and response, but by itself it does not tie that proof to the server you meant to talk to. An attacker in the middle can take a login meant for them and replay it to another server in real time. That is NTLM relay. On its own it needs a victim to try to log in to the attacker; coercion techniques remove that wait by forcing servers to do it on demand.
How It Works
A relay attack has three parts: a way to get a login sent to the attacker, a relay tool (such as Impacket's ntlmrelayx) that forwards it, and a target service that accepts it.
| Coercion method | Windows interface it abuses | Notes |
|---|---|---|
| PetitPotam | MS-EFSR (encrypting file system) | Unauthenticated form patched in 2021 (CVE-2021-36942); authenticated form often still works |
| PrinterBug (SpoolSample) | MS-RPRN (print spooler) | Works wherever the Print Spooler service runs, including many domain controllers |
| DFSCoerce | MS-DFSNM (distributed file system namespaces) | Published 2022; needs a domain account |
| ShadowCoerce | MS-FSRVP (file server VSS agent) | Only where the File Server VSS Agent is installed; patched in CVE-2022-30154 |
| CheeseOunce | MS-EVEN (event log) | Another authenticated coercion path |
| Poisoning (Responder, mitm6) | LLMNR, NBT-NS, mDNS and IPv6 DHCP | Captures logins from users and machines on the local network without coercing a server |
Tools such as Coercer try every known method against a target in one run. Coercion can also be steered to arrive over HTTP through the WebClient service, which matters because HTTP logins can be relayed to LDAP.
| Relay target | Weakness that allows it | Impact |
|---|---|---|
| AD CS web enrollment (ESC8) | NTLM accepted over HTTP without Extended Protection | A certificate as the relayed account, often a domain controller: full domain compromise |
| AD CS RPC (ESC11) | CA does not enforce encrypted requests | Same as ESC8, over RPC |
| LDAP | LDAP signing and channel binding not enforced | Create accounts, grant delegation rights (resource-based constrained delegation) or add shadow credentials |
| SMB | SMB signing not required | Code execution on servers where the relayed account is an administrator |
| MSSQL and other services | NTLM accepted without protections | Varies; often data access or code execution |
Why It Matters
Relay turns an ordinary foothold, one domain user or one compromised laptop, into control of servers and often the whole domain, without cracking a single password. It is one of the most common critical findings in internal penetration tests, because the defaults on older Windows networks allow it and the fix spans several teams (identity, PKI, servers and network).
How to Test for It
- Find where SMB signing is not required and where LDAP signing and channel binding are not enforced.
- Check AD CS web enrollment and RPC endpoints for ESC8 and ESC11.
- Identify which coercion interfaces respond on domain controllers and other high-value servers, and whether the Print Spooler and WebClient services are running.
- Check whether LLMNR, NBT-NS and IPv6 DHCP poisoning capture logins on user networks.
- In an authorized test, chain a coercion method to a relay target to confirm real impact, then clean up anything created.
- Check whether relay and coercion activity raised alerts: unusual machine-account logins from new sources and certificate requests from domain controllers.
How to Reduce the Risk
Close the targets first, because coercion methods keep being discovered: require SMB signing on all servers and clients (newer Windows releases require it by default), enforce LDAP signing and channel binding on domain controllers, require HTTPS with Extended Protection for Authentication on AD CS web enrollment or remove it, and enforce encrypted requests on the CA. Then reduce coercion: disable the Print Spooler on domain controllers and servers that do not print, apply RPC filters for the EFSR and DFSNM interfaces where they are not needed, and disable LLMNR and NBT-NS and handle IPv6 DHCP properly. Finally, audit NTLM use and restrict it, starting with domain controllers and privileged accounts (Protected Users).
Regulatory and Framework Context
For US organizations, these controls appear in CISA and NSA Active Directory hardening guidance and in CIS Benchmarks for Windows Server. Relay and coercion are tested in every internal penetration test and Active Directory assessment we run.
Latest intel on this topic
- Attackers Exploit UnrealIRCd Flaw That Enables Server-Level Command InjectionThe Hacker News · November 14, 2026
- Inside an AI Attack Against a US RetailerDark Reading · November 12, 2026
- Critical Rails Bug Threatens to Derail Vulnerable ServersDark Reading · November 4, 2026
- ROPC - So, you think you have MFA?Embrace The Red · October 20, 2022
- Beware of the Shadowbunny! at BSides SingaporeEmbrace The Red · August 28, 2020
Updated hourly from the daily briefing.
Common questions
What is PetitPotam?
PetitPotam abuses a Windows file encryption interface (MS-EFSR) to make a server, including a domain controller, authenticate to a machine the attacker chooses. Combined with relay to AD CS web enrollment (ESC8) it was a direct path to domain admin. Microsoft patched the unauthenticated form in 2021 (CVE-2021-36942), but authenticated coercion through this and other interfaces still works in many networks.
What is the difference between PetitPotam and DFSCoerce?
They use different Windows interfaces to force a login: PetitPotam uses the encrypting file system protocol (MS-EFSR) and DFSCoerce uses the distributed file system namespace protocol (MS-DFSNM). Blocking one does not block the other, which is why fixing the relay targets matters more than chasing each coercion method.
Does SMB signing stop relay attacks?
Required SMB signing stops relay to SMB on that server. It does not stop relay to LDAP, AD CS web enrollment or other services, so LDAP signing and channel binding and Extended Protection on web services are needed as well.
Should we disable NTLM?
Removing NTLM is the long-term fix, and Microsoft has announced it is phasing NTLM out. Most organizations need to audit NTLM use first, because older applications and devices still depend on it, then restrict it step by step.
Is Kerberos immune to relay?
Not entirely. Kerberos relay is possible in some configurations, which is one more reason signing and channel binding should be enforced on LDAP and other services.
Change log
- November 14, 2026First published.