Sunday, November 15, 2026

Coercing NTLM over HTTP.sys: inside AdminToSystem-HTTP.sys, a Windows admin-to-SYSTEM PoC

Coercing NTLM over HTTP.sys: inside AdminToSystem-HTTP.sys, a Windows admin-to-SYSTEM PoC

A proof of concept demonstrating how a local HTTP.sys listener can capture an NTLM token from a coerced SYSTEM process, documented for Windows privilege-escalation research and authorized assessments.

Toolivancabrera02/AdminToSystem-HTTP.sys — C++ proof of concept escalating from Windows Administrator to SYSTEM via NTLM coercion against a local HTTP.sys listener
CategoryWindows local privilege escalation / authentication coercion research
Primary UseDemonstrating, in authorized lab environments, why an Administrator token is not equivalent to SYSTEM, by coercing SYSTEM-context authentication and capturing the resulting NTLM material
Safe UseEducational analysis for authorized penetration testers and defensive researchers working exclusively on systems they own or have written permission to test; not for use against third-party infrastructure
Telemetry NoteIts mechanics depend on NTLM authentication traffic toward a local http:// listener, which defenders can observe through Windows Event Log security auditing (4624 logon events with NTLM packages), Sysmon network and process events, and Network NTLM logging

There is a persistent misconception in Windows operations that an Administrator account and the SYSTEM account are functionally interchangeable, and AdminToSystem-HTTP.sys from ivancabrera02 exists largely to refute that assumption in the most concrete way possible. The repository, written in C++, is a proof of concept that walks the last mile of local privilege escalation on Windows: taking a session that already holds administrative rights and converting it into full SYSTEM context. The mechanism it documents is not a memory-corruption bug or an unpatched zero-day; it is an abuse of legitimate Windows authentication plumbing, which is precisely what makes it interesting to both offensive and defensive practitioners studying under an authorized scope.

The core idea, as the project's summary describes it, is authentication coercion. The tool stands up a listener backed by HTTP.sys, the kernel-mode HTTP protocol stack that ships with Windows and underpins services like IIS. When a process running as SYSTEM is induced to make an HTTP request to that local listener, the HTTP.sys-hosted endpoint negotiates authentication, and the SYSTEM client obliges by presenting its NTLM credentials. The listener side then captures that token material. In other words, the tool weaponizes a trust assumption: SYSTEM services authenticate to endpoints they are pointed at, and whoever controls the endpoint controls the proof of identity that arrives.

Once the NTLM token from a SYSTEM process is in hand, the escalation completes by using that captured material to spawn a SYSTEM-context shell. The conceptual chain is therefore: host an authenticating HTTP.sys listener locally, coerce a SYSTEM process into authenticating against it, capture the token, and reinstantiate execution at the higher privilege. It is worth stressing for readers approaching this educationally that each link in that chain is a well-studied Windows behavior, not a novel flaw; the PoC's contribution is packaging them into a single demonstration of why the Administrator-to-SYSTEM boundary is softer than many hardening guides assume.

For red teamers operating under explicit authorization, tools like this fill a genuine gap in tradecraft. Many post-exploitation scenarios deliver an administrative but non-SYSTEM foothold, for example through a service account in the local administrators group, and the question becomes how to reach SYSTEM cleanly. Classical techniques rely on named-pipe impersonation, token theft with tools like incognito-style approaches, or service creation. A coercion-based path through HTTP.sys is notable because it exercises the kernel HTTP stack rather than the more commonly audited service control manager path, which can matter when the assessment goal is to evaluate detection coverage across the full range of documented escalation primitives.

The defensive value of studying this project is at least as high as the offensive value. Authentication coercion attacks against local and network listeners are a family, and this PoC gives blue teams a concrete, self-contained artifact for validating detection logic in a lab. The observable surface is meaningful: coercing authentication generates NTLM traffic that is visible in Windows Event Log security events, particularly logon events logged with NTLM negotiation rather than Kerberos, and Sysmon can correlate the process that made the outbound connection with the process hosting the listener. A local SYSTEM shell appearing immediately after unusual NTLM activity on loopback is exactly the kind of sequence a tuned detection engineering program should surface.

Hardening guidance follows directly from the mechanism. Environments that have disabled or tightly constrained NTLM through Group Policy, moving authentication to Kerberos where possible, structurally reduce the value of captured NTLM material. Endpoint protection that flags unsigned or unrecognized binaries registering HTTP.sys URL prefixes — which requires administrative rights to do — adds another layer of friction. The deeper lesson for defenders is monitoring identity context, not just binary reputation: any escalation path that relies on one process handing credentials to another is blunted when credential presentation is itself an audited, alertable event.

The repository itself is modest, with 49 stars at the time of this writing, no license file declared, and a README that is sparse or absent, with the project summary written in Spanish describing the escalation from Administrador to SYSTEM. That thinness means prospective users must read the C++ source on the main branch to understand exactly which HTTP.sys APIs are invoked and how the token capture is implemented, which is arguably the right posture anyway: anyone evaluating an escalation PoC should audit the code before running it anywhere, and the absence of packaging polish is a reminder that this is research code, not a product. The absence of a license also means default copyright restrictions apply, so organizational use policies should be checked before internal redistribution.

Where this tool fits in the broader landscape is alongside other modern admin-to-SYSTEM demonstrations that lean on Windows authentication internals rather than kernel exploits — the general family of token impersonation and coercion techniques popularized by tooling like Potato-variant escalations, which similarly trick privileged accounts into authenticating to attacker-controlled listeners. AdminToSystem-HTTP.sys distinguishes itself by using the native kernel HTTP stack as the listener substrate, which keeps the footprint native to Windows rather than relying on custom sockets. For students of Windows internals, comparing how HTTP.sys handles authentication negotiation versus a userspace HTTP server is itself an instructive exercise in where trust boundaries actually live inside the OS.

Used responsibly, this is a valuable educational artifact: it makes the Administrator-versus-SYSTEM distinction tangible, gives authorized assessors a reproducible primitive to test against their own hardened builds, and gives defenders a clear set of telemetry anchors — NTLM logon events, Sysmon network connections, and HTTP.sys namespace registrations — to build detections around. It should go without saying, and bears repeating anyway, that running credential-capture tooling against any system you do not own or administer is both illegal and outside every legitimate engagement scope; the analysis here is documentary, intended for professionals working in labs and authorized engagements who want to understand, and ultimately detect, this class of escalation.

Official project repository for ivancabrera02/AdminToSystem-HTTP.sys.
Download Tool

Educational analysis for authorized security professionals. Use only in controlled, authorized environments.

Sponsored
Share articleFacebookXLinkedIn

Continue exploring

Browse all articles →

0 comentários:

Post a Comment

Note: Only a member of this blog may post a comment.