Join our Newsletter — 33% off our NHI Course
Home› Guides› MITRE ATT&CK T1127 Trusted Developer Utilities Proxy Execution:…
Guide SOC & Detection Engineering

MITRE ATT&CK T1127 Trusted Developer Utilities Proxy Execution: Detecting Evasion

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 12 November 2026 11 min read
On this page

Trusted Developer Utilities Proxy Execution (T1127) is the abuse of signed build tools, debuggers and deployment frameworks to run attacker code through a process that application control and reputation checks already trust. In ATT&CK v20 it sits under two tactics, Stealth (TA0005) and Execution (TA0002), on Windows, with three sub-techniques: MSBuild (T1127.001), ClickOnce (T1127.002) and JamPlus (T1127.003). The core problem is that developer tools are designed to compile and run code, so any machine that has them and does not need them carries a ready-made execution engine. This page covers the sub-techniques, recorded cases, the v20 detection strategies, the mitigations ATT&CK maps (M1038, M1042, M1021, M1045) with concrete settings, and how to test. Windows system binaries such as rundll32 and mshta are covered in the T1218 System Binary Proxy Execution Defensive Guide; this page stays on developer utilities.

Key takeaways

  • Developer tools on non-developer machines are the root cause. Microsoft's App Control recommended block list includes msbuild.exe, the Windows debuggers (cdb.exe, windbg.exe, kd.exe), dotnet.exe, csi.exe, rcsi.exe and fsi.exe unless a scenario explicitly needs them.
  • Reputation is not intent. Elastic Security Labs showed in August 2024 that "reputation hijacking" of trusted programs such as JamPlus can bypass Windows Smart App Control without warnings.
  • State actors use it in current campaigns. ESET recorded MirrorFace using MSBuild to compile and run its FaceXInjector tool against a Central European diplomatic institute in August 2024.
  • ClickOnce trust prompts are a policy setting. The PromptingLevel registry key controls whether ClickOnce apps from the Internet zone can be installed at all.
  • Detect by context. ATT&CK v20 detection focuses on a developer utility running outside a developer or build context, then spawning shells, loading unsigned modules, writing executables or connecting out.

What T1127 is and how the sub-techniques differ

ATT&CK describes T1127 as using utilities for software development, debugging and reverse engineering to proxy execution of malicious payloads. These tools are often signed with legitimate certificates, so they run where application control trusts the signer, and the payload runs inside them. ATT&CK also notes that Smart App Control, which checks unsigned applications against a Microsoft cloud service, can be bypassed by abusing signed applications that support running arbitrary code.

Sub-techniqueMechanism (per ATT&CK v20)Typical abuseATT&CK mitigations
T1127.001 MSBuildMSBuild.exe processes XML project files; inline tasks (introduced with .NET 4) let C# or Visual Basic code be embedded, compiled and runProject file with an inline task loads shellcode or a .NET loaderM1038 Execution Prevention, M1042 Disable or Remove Feature or Program
T1127.002 ClickOnce.application and .appref-ms deployments install self-updating .NET apps from a share or web page with minimal interaction and no admin rightsMalicious app disguised as legitimate software; launch through rundll32 with dfshim.dll; startup folder persistenceM1042, M1021 Restrict Web-Based Content, M1045 Code Signing
T1127.003 JamPlusJamPlus build utility runs scripts described in .jam build filesScript run from a reputable build tool to slip past Smart App ControlM1038, M1042

The parent technique also covers other developer utilities named by ATT&CK's references, such as the .NET DNX and rcsi.exe hosts, the WinDbg and CDB debuggers, and Tracker.exe from the LOLBAS project.

Real cases

Actor or softwareWhat was recordedDefender lesson
MirrorFace (G1054), Operation AkaiRyū (C0060)Used MSBuild to compile and run FaceXInjector, a C# injection tool stored in an XML file, which injected the HiddenFace backdoor into a legitimate Windows utility; compiled on each scheduled task run (ESET, March 2025; activity from August 2024)MSBuild launched by a scheduled task on a non-developer host is a high-confidence alert
NOOPLDR (S9025)Can be executed via MSBuild (ATT&CK, citing JPCERT/CC, July 2024, on MirrorFace)Loader families reuse the same path across campaigns
PlugX (S0013)A version loads as shellcode within a .NET Framework project using msbuild.exe, presumably to bypass application control (Palo Alto Networks, 2017)Long-running technique, still effective where MSBuild is allowed
Frankenstein campaign (C0001)Threat actors used MSBuild to execute an actor-created file (Cisco Talos, June 2019)Combined with document lures for initial execution
Empire (S0363)Built-in modules abuse trusted utilities like MSBuild.exeOpen-source frameworks make it routine

See the MirrorFace Threat Actor Profile Guide and the Mustang Panda Threat Actor Profile Guide for PlugX context.

What to detect and where

ATT&CK v20 gives the parent and each sub-technique its own detection strategy. All of them are Windows behaviour chains.

Strategy (analytic)Behaviour to detectLog sources named in v20
DET0172 (AN0488), parentA signed developer utility runs in a non-developer context and spawns powershell.exe, cmd.exe, rundll32.exe, regsvr32.exe or wscript.exe, loads unsigned or user-writable DLLs, writes then runs a new PE from a user-writable path, or connects out immediatelySecurity 4688; Sysmon 3, 7, 11, 22; AppLocker audit and block events
DET0556 (AN1535), MSBuildMSBuild.exe outside build contexts or with anomalous arguments (remote shares, non-canonical paths, encoded property values), followed by child LOLBins, new artefacts, unsigned module loads, injection or network connectionsSecurity 4688; Sysmon 3, 7, 8, 10, 11, 22; CodeIntegrity Operational; AMSI where available
DET0191 (AN0550), ClickOncerundll32.exe invoking dfshim.dll with ShOpenVerbApplication, or dfsvc.exe spawning unexpected children or loading unsigned modulesSecurity 4688; Sysmon 7; Security-Mitigations telemetry
DET0585 (AN1610), JamPlusJamPlus.exe launching payloads through crafted .jam files, with abnormal process creation or artefacts outside development workflowsSecurity 4688; Sysmon 3, 11, 22; CodeIntegrity Operational

Building context

  • Define developer and build hosts: keep an inventory (device group or tag) of developer workstations and build agents. MSBuild, debuggers and dotnet on any other host should be rare enough to alert on directly.
  • Baseline parents on build hosts: on build agents, MSBuild is normally launched by the CI agent service, Visual Studio or dotnet. MSBuild launched by Office, a browser, explorer.exe with a project file from Downloads, or a scheduled task is suspicious even there.
  • Watch project file locations: .csproj, .proj or .xml project files in AppData, Temp, Public or network shares, especially newly written ones, are a lead.
  • Network from a compiler: MSBuild restoring packages talks to your package feeds; MSBuild connecting to an unknown host is not package restore.
  • ClickOnce sources: dfsvc.exe activity that follows a browser download of an .application file from an Internet site is worth a look in any estate that does not use ClickOnce.

Mitigations with concrete settings

MitigationWhat to doSource
M1038 Execution PreventionApply Microsoft's recommended block rules in App Control for Business (WDAC): they deny msbuild.exe, msbuild.dll, Microsoft.Build.dll, the debuggers (cdb, windbg, kd, ntkd, ntsd, dbghost, dbgsvc, dbgsrv), dnx.exe, dotnet.exe, rcsi.exe, csi.exe, fsi.exe, texttransform.exe and others. Microsoft says to allow msbuild.exe only where the reference system is used to build managed applicationsMicrosoft Learn, Applications that can bypass App Control
M1042 Disable or Remove Feature or ProgramDo not install Build Tools, SDKs, debuggers or JamPlus on standard user, server or kiosk images; remove them where presentATT&CK v20
M1021 Restrict Web-Based ContentSet HKLM\SOFTWARE\MICROSOFT\.NETFramework\Security\TrustManager\PromptingLevel, value Internet, to Disabled so only apps signed with an explicitly trusted certificate install from the Internet zone (Microsoft's default for applications in that zone is Enabled)Microsoft Learn, Configure the ClickOnce trust prompt behavior
M1045 Code SigningRequire Authenticode-signed ClickOnce deployments from trusted publishers (AuthenticodeRequired or Disabled on other zones as your use allows)ATT&CK v20; Microsoft Learn

The App Control block list is written for allow-list policies; deploy it in audit mode first, review CodeIntegrity Operational events for legitimate use, then enforce. See the App Control for Business (WDAC), the M1038 Execution Prevention, the M1042 Disable or Remove Feature or Program, the M1021 Restrict Web-Based Content and the M1045 Code Signing.

Decision guidance: developer machines

Blocking is easy on standard endpoints. Developer workstations and build agents need the tools, so the choice is about containment rather than removal.

  • Standard user and server estate: enforce the block list. There is rarely a business reason for MSBuild or a debugger on a finance laptop or a domain controller.
  • Developer workstations: allow the tools but monitor parent processes and project file locations, and keep developers on separate device groups with their own policy and alerting thresholds.
  • Build agents: allow MSBuild only for the CI agent service account and expected parents, use ephemeral agents where possible, and alert on interactive logons.
  • Smart App Control alone is not enough: Elastic's research shows reputation-based checks can be bypassed by trusted programs that run scripts, so pair it with explicit deny rules where you manage policy.

Where identity fits

Developer machines and build agents hold some of the most valuable non-human credentials in an organisation: package feed tokens, signing keys, cloud deployment credentials and CI service account secrets. Code run through MSBuild on a build agent runs with the agent's identity and can read those secrets. Scope CI service accounts to the minimum, keep signing keys in hardware or a managed signing service, and rotate build secrets after any suspicious MSBuild activity on an agent. See the Developer Workstation and IDE Security Guide.

How to test

  1. Inventory: query EDR or software inventory for msbuild.exe, the debuggers, dotnet.exe and JamPlus on non-developer hosts; that list is your exposure.
  2. Audit policy: deploy the recommended block rules in audit mode to a pilot ring and review would-be blocks for a few weeks.
  3. Emulate: with an authorised emulation framework, run a benign MSBuild inline task that starts a harmless child process on a test host outside the developer group.
  4. Verify detection: confirm DET0556-style alerts fire with parent, project path and child process.
  5. Enforce and verify block: move the pilot to enforcement and confirm the same test is blocked and logged.
  6. Rollback plan: keep the previous policy version ready; App Control policies can be replaced through the same management channel if a business tool breaks.

Measures: non-developer hosts with developer utilities installed (target zero), share of estate with the block list enforced, MSBuild executions outside the developer and build groups per month, and time to alert in tests.

Common failure patterns

  • Allow by signer: trusting everything signed by Microsoft lets MSBuild and the debuggers through, which is the point of the technique.
  • Golden images with SDKs: build tools installed on every image "in case" turn every endpoint into a compiler.
  • No developer device group: without one, rules cannot tell normal builds from abuse and get tuned into silence.
  • Default ClickOnce prompts: leaving the Internet zone at Enabled lets users install unsigned ClickOnce apps after a single prompt.
  • Treating build agents as trusted: agents run untrusted code by design and hold valuable secrets, so they need monitoring and short-lived credentials.

Framework mapping

FrameworkReferenceRelevance
MITRE ATT&CK v20T1127, T1127.001, T1127.002, T1127.003; TA0005, TA0002Technique, sub-techniques and tactics
MITRE ATT&CK v20M1038, M1042, M1021, M1045; DET0172, DET0556, DET0191, DET0585Mapped mitigations and detection strategies
MITRE ATT&CK v20T1218, T1218.011, T1204, T1547.001, T1055Related techniques named in the sub-technique descriptions and cases
MITRE D3FENDD3-EAL Executable Allowlisting, D3-EDL Executable Denylisting, D3-PLA Process Lineage Analysis, D3-PSA Process Spawn AnalysisBlock lists and parent-child analysis
NIST SP 800-53 Rev 5CM-7(2), CM-7(5), CM-11, SI-4, SI-7(15)Prevent program execution, allow-by-exception, user-installed software, monitoring, code authentication
NIST CSF 2.0PR.PS-05, PR.PS-01, DE.CM-09Unauthorised software prevention, configuration management, endpoint monitoring

Practitioner checklist

  • Inventory developer utilities on non-developer hosts and remove them.
  • Deploy Microsoft's App Control recommended block rules in audit, then enforce.
  • Create developer and build agent device groups with their own policy and alert thresholds.
  • Alert on MSBuild with unusual parents, project files in user-writable paths or outbound connections.
  • Set the ClickOnce Internet zone PromptingLevel to Disabled where ClickOnce from the Internet is not needed.
  • Alert on rundll32 with dfshim.dll and on dfsvc.exe spawning unexpected children.
  • Collect CodeIntegrity Operational and AppLocker events centrally.
  • Scope and rotate build agent credentials and signing keys.

Standards and references

Related NHI Mgmt Group resources: Living off the Land Guide · ATT&CK Stealth and Defense Impairment Tactics Guide · ATT&CK Execution Tactic Guide · Application Control and Endpoint Prevention Buyer's Guide · Sysmon Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 12 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.