Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
CVE-2026-97798_LPE_Windows — CVE-2026-97798 PoC — WER Time Travel Debugging argument injection to SYSTEM (Windows 11 LPE), with an adaptive faultrep.dll resolver. | Kitploit
Tools/GitHubGitHub/j0wii/cve-2026-97798_lpe_windows
Privilege EscalationVulnerability AnalysisExploitationReverse EngineeringBinary AnalysisRed Teaming
GitHubj0wii/cve-2026-97798_lpe_windows

CVE-2026-97798_LPE_Windows

Share

CVE-2026-97798 PoC — WER Time Travel Debugging argument injection to SYSTEM (Windows 11 LPE), with an adaptive faultrep.dll resolver.

View Repository
136820 days agoNot yet reviewed

CVE-2026-97798 — WER Time Travel Debugging argument injection to SYSTEM

Authorized laboratory use only. This repository is for vulnerability research, exploit validation and detection engineering inside disposable Windows virtual machines that you own or are explicitly authorized to test. See LICENSE.

An unprivileged local caller can preserve attacker-controlled command-line syntax across the Windows Error Reporting (WER) privilege boundary. The elevated WER collector later formats that string into a TTTracer.exe command line without preserving argument boundaries, so injected -out / -launch switches are interpreted by the Time Travel Debugging recorder and an attacker-selected executable is launched with the collector's elevated token.

The result is a local privilege escalation from a standard user (and from a UAC-filtered administrator) to NT AUTHORITY\SYSTEM.

low / medium-integrity process
  -> faultrep.dll!CElevatedDataCollection::ControlTimeTravelDebugging
  -> faultrep.dll!SpawnDataCollector
  -> Windows Error Reporting service (elevated collector)
  -> CDataCollection::ControlTimeTravelDebugging
  -> TtdControl::LaunchRecordingProcess
  -> TTTracer.exe / TTDRecord.dll command-line parser
  -> attacker-controlled -launch target

This is an argument injection issue (CWE-88), not a shell command injection: no cmd.exe /c is involved at the vulnerable sink.

Repository layout

PathDescription
bin/CVE-2026-97798.exeAdaptive PoC, lean build: exploit, --resolve, --stop-ttd
bin/CVE-2026-97798-diag.exeSame code plus --diagnose and --sweep troubleshooting modes
src/CVE-2026-97798.cppSingle source for both binaries (-DCVE97798_LEAN selects the lean build)
build.ps1Builds both binaries with LLVM-MinGW or MinGW-w64 and prints SHA-256
docs/TROUBLESHOOTING.mdRejections such as 0x80070032, and how to tell them apart
README.es.mdSpanish version of this document

Both binaries are one self-contained x64 image (statically linked, no runtime dependency) that re-enters itself in several internal modes. At runtime it stages two aliases of that same image under C:\Users\Public\Documents\CVE97798\: one for the integrated console client and one for the elevated payload. No extra DLL, script or cmd.exe is required.

Quick start

Use a disposable snapshot of a vulnerable build. From a non-elevated PowerShell or Command Prompt — the PoC must not run elevated:

# 1. optional: try to clear leftover recorder state from an earlier attempt
#    (see "Recorder cleanup" below: without admin this only reports what is stuck)
.\bin\CVE-2026-97798.exe --stop-ttd

# 2. run it; on success you get an interactive SYSTEM shell
.\bin\CVE-2026-97798.exe

Expected output of the integrated console:

[CVE-2026-97798] Exploit shell token:
nt authority\system

Machine-readable proof is written to:

C:\Users\Public\Documents\CVE97798\system-proof.txt

Examples of the proof file from a filtered-administrator run:

cve=CVE-2026-97798
user=NT AUTHORITY\SYSTEM
is_system=true
is_elevated=true
payload_session=0
shell_session=0
spawn_error=0

Do not run it twice in a row

A second execution fails with HRESULT 0x80070032 while the recorder from the first run is still alive. Close the PoC console, clean up (see Cleanup), then retry.

Modes

CommandBinaryEffect
(no arguments)bothFull exploit: integrated console + elevated payload + SYSTEM shell
--markerbothOnly write system-proof.txt, no interactive shell
--external-consolebothPayload without the integrated console client
--resolve <dll>bothResolver self-test against an offline faultrep.dll; sends nothing
--stop-ttdbothBest-effort recorder cleanup: terminates what it can, empties the trace directory, and reports exactly why a leftover recorder survived (usually: it is owned by SYSTEM)
--cleanup-systembothRemove elevated leftovers of the PoC itself
--diagnosediagPrecondition report: versions, live recorders, trace leftovers, TTD/WER policy, collector vtable
--sweepdiagReplay 7 request shapes through ControlTimeTravelDebugging, printing each HRESULT

Exit codes: 0 success · 2 usage · 3/4/5 lab preparation · 40/41 COM/faultrep.dll · 42 resolver rejected the image · 43 SpawnDataCollector failed · 44 CoSetProxyBlanket · 45 request rejected by ControlTimeTravelDebugging · 46 marker is not SYSTEM · 47 console terminated · 48 marker timeout · 49 DLL not mappable · 50 no unique resolver candidate · 51/52 TTTracer failure in --cleanup-system.

The adaptive resolver

The binaries embed no build-specific RVA. At runtime the loaded faultrep.dll is parsed through its own PE64 exception table (IMAGE_DIRECTORY_ENTRY_EXCEPTION), and every function containing the WER collector opcode is scored on four protocol invariants:

InvariantValuePoints
collector opcode, exactly once0xD00000022
request/message size immediate0x578 (1..4 occurrences)1
recorder wait timeout10000 ms (1..2 occurrences)1
reply marker, exactly once0xD00000001

A function is only accepted at 5/5, inside a plausible size band (0x300..0x2000), and only if it is the unique accepted candidate in the image. Anything else is rejected instead of being called speculatively, so an unseen or shifted build produces a clean refusal (exit 42/50) rather than a crash. The score profile is printed so the shape of the build is visible:

.\bin\CVE-2026-97798.exe --resolve C:\Windows\System32\faultrep.dll
# SpawnDataCollector RVA=0x4498c profile=multi-header score=5/5

multi-header means the build keeps two 0x578 immediates (two request paths); single-header means one. Both are accepted.

Validated environment

ComponentValue
OSWindows 11 24H2 x64, build 26100.9550
Vulnerable faultrep.dll10.0.26100.9549 (resolver RVA 0x4CEB0)
Additional validated buildWindows 11 22H2, build 22621.525, faultrep.dll 10.0.22621.1408 (resolver RVA 0x4498C, profile multi-header)
Fixed comparisonfaultrep.dll 10.0.26100.9711 (resolver RVA 0x4E40C)
TTTracer / TTDRecord1.11.580.0

Windows 11 22H2 is out of support and is not listed among the affected products in MSRC guidance; it still contains the vulnerable collector, but the recorder shipped with it may not accept every request shape. See docs/TROUBLESHOOTING.md before assuming a failure is the PoC's fault.

Building

Requirements: an x64 toolchain able to produce a static PE, either LLVM-MinGW or MinGW-w64.

.\build.ps1

# or point at the compiler explicitly
.\build.ps1 -Compiler C:\toolchains\llvm-mingw\bin\x86_64-w64-mingw32-clang++.exe

The script builds the lean binary with -DCVE97798_LEAN, builds the diagnostic binary from the same source, strips symbols when a strip tool is available, and prints SHA-256 for both. Builds are static and suppress the PE linker timestamp, so identical source plus toolchain produce an identical hash.

Manual equivalent:

$flags = "-std=c++20 -O2 -static -municode -Wall -Wextra -Wno-unused-parameter -Wl,--no-insert-timestamp"
$libs  = "-lole32 -loleaut32 -ladvapi32 -lversion"

x86_64-w64-mingw32-g++ $flags.Split() -DCVE97798_LEAN src/CVE-2026-97798.cpp -o bin/CVE-2026-97798.exe $libs.Split()
x86_64-w64-mingw32-g++ $flags.Split() src/CVE-2026-97798.cpp -o bin/CVE-2026-97798-diag.exe $libs.Split()

Cleanup

Download Tool

Most Popular

View all →

Discover the most used tools by our community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 The ultimate OSINT toolkit: investigation graph, confidence scoring, alias families, map, timeline and reports. 49 modules, 400+ sites, no API keys required.

    786
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    GhostLock One-Tap Execution App (CVE-2026-43499)

    506
  • 3

    ARTEX

    baarai/artex

    LLM multi-agent autonomous penetration testing platform with planner/worker agents, asset and exploration graphs, MITM traffic recording, and vulnerability retesting.

    429
  • 4

    redforge

    cybersecurity91/redforge

    Browser-side red team toolkit with 102 offline tools for recon, payload generation, credential attacks, crypto, encoding and engagement reporting.

    345
  • 5

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    The system of action for AI-native cybersecurity—where intent becomes governed execution, evidence becomes operational memory, and every operation improves the next.

    262
  • 6

    looking-glass

    anonghosty/looking-glass

    CSIRT platform that ingests Shadowserver reports into MongoDB and replays them as interactive maps, analysis, findings, advisories and executive dashboards.

    256

Explore all tools

Browse our collection of tools

View all tools →