NTLM relay attacks do not require the attacker to crack a single password. They relay the victim's own authentication response in real time, and 95% of enterprise Active Directory environments still have NTLM enabled on at least one domain-joined system according to CrowdStrike's 2025 Identity Threat Report. The result: a credential attack that bypasses every password-strength control an organisation has ever implemented.
An NTLM relay attack (MITRE ATT&CK T1557.001) is a man-in-the-middle credential attack where an attacker intercepts an NTLM authentication challenge from a victim host, relays it in real time to a target server, and authenticates to that server without possessing the underlying password. The attack chain is entirely passive until the relay step. No malware is deployed. No brute force runs. The victim authenticates themselves to the attacker's chosen target.
The standard NTLM relay chain has four steps, each of which uses Windows features that exist in every default domain configuration.
Step one is name resolution poisoning. Windows broadcasts Link-Local Multicast Name Resolution (LLMNR) queries and NetBIOS Name Service (NBT-NS) queries when DNS fails to resolve a hostname. An attacker running Responder or Inveigh on the same network segment listens for these broadcasts. When a workstation asks "is there a host named FILESERV01?" and DNS has no answer, Responder replies first: "Yes, that is me." The victim's machine accepts this answer.
Step two is credential capture. The victim's machine, believing it is talking to the requested host, initiates NTLM authentication. It sends an NTLM Negotiate message. The attacker's tool forwards that to the actual target server and relays the challenge back to the victim. The victim signs the challenge with its NTLM hash and sends the response. The attacker now has a valid, signed NTLM authentication exchange.
Step three is relay. The attacker forwards that exchange to the target server immediately, before the exchange expires. If SMB signing is not required on the target, the server accepts the authentication. The attacker now has an authenticated session on the target system as the victim user.
Step four is exploitation. With an authenticated session, the attacker can read shares, execute commands via Service Control Manager (if the victim account has admin rights), or dump credentials from the target. If the victim was a domain administrator, the attacker has full domain access.
Responder (by Laurent Gaffie, available on GitHub) handles the poisoning phase. It is preinstalled in Kali Linux and runs as a single command. It poisons LLMNR, NBT-NS, and MDNS simultaneously. The setup takes under two minutes for an attacker already on the network segment.
Impacket's ntlmrelayx.py handles the relay phase. It accepts the credential capture from Responder, relays it to a specified target or list of targets, and executes commands if the victim has administrative rights. A common combination is Responder with --no-smb-server flag (so Responder does not handle SMB itself) forwarding to ntlmrelayx, which runs the actual relay against target systems.
Inveigh is a PowerShell-based alternative that runs entirely in memory. It performs the same poisoning and optionally relays, making it harder to detect through traditional file-based scanning.
Microsoft formally announced the deprecation of NTLM in Windows 11 24H2 in September 2023. Deprecation is not removal. NTLM remains enabled by default to preserve compatibility with legacy applications, printers, scanners, and industrial control systems that cannot use Kerberos or modern authentication methods. In a 2025 survey of enterprise environments, Mandiant found NTLM traffic on internal networks in 89% of assessments, even in organisations that believed they had disabled it.
Three factors sustain the attack's effectiveness. First, LLMNR and NBT-NS are still enabled by default in Windows. Second, SMB signing was enforced by default only on domain controllers, not workstations, until Windows 11 build 25381 in 2023. Organisations running older Windows versions or non-default configurations remain exposed. Third, the attack requires only network access. An attacker who compromises a single VPN credential, reaches a guest Wi-Fi segment connected to the corporate network, or pivots from an IoT device has the access they need.
CISA explicitly lists LLMNR/NBT-NS poisoning as a technique used in ransomware precursor activity in its 2024 advisory AA24-038A, noting that "adversaries use this technique to move laterally from initial access to domain-privileged accounts within hours."
Detection requires both network-level and event log monitoring. Neither source alone catches all relay activity.
LLMNR operates on UDP port 5355. NBT-NS operates on UDP port 137. Capturing these protocols on a network tap or span port reveals two patterns. Any host broadcasting LLMNR queries for hostnames that do not exist in DNS is producing the raw material for a poisoning attack. Any host responding to LLMNR or NBT-NS queries that is not a domain controller or DNS server is either misconfigured or running Responder.
Network detection rule (Suricata/Snort compatible):
alert udp any any -> 224.0.0.252 5355 (
msg:"LLMNR Query Detected - Potential NTLM Relay Setup";
sid:9000001; rev:1;
)
An alert is not an incident. LLMNR traffic exists legitimately in many environments. The signal is an unexpected host answering LLMNR queries rather than the domain controller.
| Event ID | Log | Relay Indicator |
|---|---|---|
4624 |
Security | Logon Type 3 (network) from an unexpected source IP for a privileged account |
4648 |
Security | Explicit credential logon to a system the account does not normally access |
4625 |
Security | Failed NTLM logon from a host that is not the account's normal workstation |
7045 |
System | New service installed on a target host (attacker used relay to create a service for code execution) |
The most reliable SIEM query combines Event ID 4624 (Logon Type 3) with a source IP that does not match the user's registered workstation. Legitimate network logons come from known sources. An authenticated session originating from a host the account has never touched before is worth immediate investigation.
Splunk implementation:
index=wineventlog EventCode=4624 Logon_Type=3
| stats values(Source_Network_Address) as src_ips by Account_Name
| where mvcount(src_ips) > 2
| mvexpand src_ips
Tune the threshold for your environment. The goal is surfacing accounts that authenticate from more source IPs than their normal pattern allows.
Navigate to: Computer Configuration > Administrative Templates > Network > DNS Client. Set "Turn off multicast name resolution" to Enabled. This disables LLMNR entirely on all domain-joined machines the policy applies to.
Apply this to all OUs containing workstations and servers. Domain controllers do not need LLMNR for name resolution; apply there as well. Test in a pilot OU first. Some older applications use LLMNR for service discovery rather than DNS. Identify those before domain-wide rollout.
NBT-NS (NetBIOS Name Service) on UDP 137 is the older protocol that Responder poisons alongside LLMNR. Disable it via DHCP scope options: set option 001 (Microsoft option) to a value of 2. This instructs DHCP clients to disable NetBIOS. For hosts with static IPs, configure it directly via Network Adapter > IPv4 Properties > Advanced > WINS tab: select "Disable NetBIOS over TCP/IP."
A Group Policy Preference can automate the registry change across all systems:
HKLM\SYSTEM\CurrentControlSet\Services\NetBT\Parameters\Interfaces\
Tcpip_{GUID}\NetbiosOptions = 2 (REG_DWORD)
Wildcard targeting of all interface GUIDs requires a startup script or PowerShell GPO preference in most environments.
SMB signing is enforced by default only on domain controllers. Workstations and member servers default to "if requested" rather than "required." An NTLM relay against a workstation or file server without required signing succeeds silently.
GPO path: Computer Configuration > Windows Settings > Security Settings > Local Policies > Security Options.
Apply to all OUs. SMB signing adds a small performance overhead. For most workloads the overhead is under 3% and undetectable in practice. For high-throughput file server traffic, benchmark before domain-wide enforcement. The alternative is leaving the attack surface open indefinitely.
Disabling LLMNR and requiring SMB signing removes the two primary attack paths. The residual risk is NTLM relay against HTTP, LDAP, and other protocols. For complete coverage, audit NTLM usage first, then restrict it.
Enable NTLM auditing: Computer Configuration > Windows Settings > Security Settings > Local Policies > Security Options > "Network security: Restrict NTLM: Audit NTLM authentication in this domain" set to "Enable all." This logs all NTLM authentication events under Event ID 8004 without blocking anything. Run for two to four weeks. Identify every system and application generating NTLM traffic.
Then set "Network security: Restrict NTLM: NTLM authentication in this domain" to "Deny all domain accounts." For applications that require NTLM, add them to the exception list via "Network security: Restrict NTLM: Add server exceptions for NTLM authentication in this domain."
Microsoft's target for Windows environments is Kerberos-only authentication. NTLM restriction moves organisations toward that goal without immediately breaking legacy applications.
| Control | What It Blocks | Priority |
|---|---|---|
| Disable LLMNR (GPO) | Removes the primary name-resolution vector Responder exploits | Critical |
| Disable NetBIOS over TCP/IP | Removes the NBT-NS vector, the secondary poisoning channel | Critical |
| Require SMB signing (GPO) | Blocks relay success against SMB targets even if poisoning occurs | High |
| NTLM audit mode | Maps all NTLM usage before restriction; identifies legacy dependencies | High |
| NTLM domain restriction | Forces Kerberos; blocks all NTLM relay paths including HTTP and LDAP | Medium (after audit) |
| SIEM rules on Event ID 4624 | Detects completed relays in progress; enables rapid response | Medium |
The attack has existed since the late 1990s. The tools are free. The setup takes minutes. Yet the three Group Policy changes that close the primary attack surface remain absent in most enterprise environments, usually because no one applied them when the domain was built and no audit has flagged them since.
Apply the LLMNR and SMB signing controls first. Run the NTLM audit in parallel. The relay attack surface closes significantly within a single change window.
An NTLM relay attack (MITRE ATT&CK T1557.001) is a man-in-the-middle credential attack where an attacker intercepts an NTLM authentication handshake from a victim host and relays it in real time to a target server, completing authentication without possessing or cracking the underlying password. Because no password cracking is involved, password complexity and length offer no protection against a successful relay.
Link-Local Multicast Name Resolution (LLMNR) broadcasts name-resolution queries to the local subnet when DNS fails. An attacker running Responder responds to these broadcasts, impersonating the requested host. The victim's machine sends its NTLM credentials to the attacker, who relays them to the real target server before the exchange expires. Disabling LLMNR via Group Policy removes this initial vector entirely.
SMB signing prevents relay attacks against SMB targets. When required on a server, it demands a cryptographic signature on every packet, which the attacker cannot produce without the actual password hash. This blocks relays against file servers and domain controllers. It does not block relay against HTTP, LDAP, or other protocols. A complete defense requires disabling LLMNR and NBT-NS in addition to requiring SMB signing across all systems.
Navigate to Computer Configuration > Administrative Templates > Network > DNS Client and set "Turn off multicast name resolution" to Enabled. For NetBIOS, configure DHCP scope option 001 to value 2 (disable NetBIOS), or set it directly on static-IP hosts via Network Adapter properties. Apply both controls together for complete name-resolution poisoning protection.