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
chrome-v8-maglev-literal-stale — CVE-2026-126910 - Google Chrome - High - Remote - Type confusion - Maglev object/array literal stale load. Chrome < 157.0.8103.78. | Kitploit
Tools/GitHubGitHub/abraxas/chrome-v8-maglev-literal-stale
Vulnerability AnalysisExploitationWeb SecurityLabs & Practice
GitHubabraxas/chrome-v8-maglev-literal-stale

chrome-v8-maglev-literal-stale

CVE-2026-126910 - Google Chrome - High - Remote - Type confusion - Maglev object/array literal stale load. Chrome < 157.0.8103.78.

View Repository
4 days agoNot yet reviewed
Share

Abraxas Labs - chrome-v8-maglev-literal-stale

chrome-v8-maglev-literal-stale

Chrome's JavaScript engine has a fast compiler called Maglev. It looked at a property on Object.prototype, decided the answer was "TRUE", and burned that into optimized code. Then the page created an object (or an array). That literal write kicked the prototype out of dictionary mode and onto a new map. Maglev never marked the literal as a write, so it kept using the old load. The next read came back "WRONG".

That is type confusion. Google shipped it as CVE-2026-126910, High, "arbitrary code inside the sandbox," fixed in Chrome 157.0.8103.78 on 11 November 2026. This pack is the stale load, not a shell.

Class: Type confusion (CWE-843) Reach: Remote, crafted web content

IDCVE-2026-126910
ClassType confusion (CWE-843)
ReachRemote (crafted HTML)
ProductGoogle Chrome / V8 Maglev
AffectedChrome before 157.0.8103.78 (Windows/Mac 157.0.8103.78/.79, Linux 157.0.8103.78)
FixedChrome 157.0.8103.78, V8 CL 8668740 (I3298cf355d6751a71f1c8e4f42a2797f2c4acbd7), M157 cherry-pick 8753123
Chrome ratingHigh
Authnone
LicenseGNU Affero GPL v3.0
Lablocal Docker, linux-arm64 Chrome for Testing

What an attacker can do

MethodSinkResult
Maglev-compile a load of Object.prototype.x1, then CreateObjectLiteral / CreateArrayLiteralMaglev IR missing CanWrite on those literalsLab-proven: vulnerable 157.0.8103.50 returns "WRONG" for both object and array cases; patched 157.0.8103.78 deopts and returns "TRUE"
Same map transition on a more interesting hidden classtyped load of the wrong representationInferred from the patch. Maglev treated a write as a pure create. Wrong map means wrong typed load
Turn the type confusion into renderer RCEV8 heapNot attempted. No forged pointers, no shell. Google rates the class as code exec inside the sandbox
Browser-process compromisesandbox escapeClosed for this bug. This is renderer/V8 only

%OptimizeMaglevOnNextCall in the lab is the in-tree regress harness (regress-575525282-2.js / -3.js). A real page would wait for Maglev's normal tier-up.

Why the literal was a write

CreateObjectLiteral and CreateArrayLiteral can put a prototype into fast mode, which changes its map. Maglev IR listed them as if they could not write. CL 8668740 (Marja Hölttä, reviewed by Darius Mercadier, V8 main@{#111225}) marks those nodes CanWrite. After the fix, Maglev bails when the prototype map changes, and the load sees "TRUE" again.

Chromium issue 574059463 stays restricted. The patch and the mjsunit files are public.

The lab

Two Chrome for Testing builds, headless, --js-flags=--allow-natives-syntax --maglev --no-turbofan.

  1. Warm h on a dictionary-mode prototype so Maglev sees x1 === "TRUE".
  2. Maglev-compile h.
  3. Fire CreateObjectLiteral or CreateArrayLiteral (the write).
  4. Read obj.x1.

Vulnerable pin: Maglev stays optimized (status 25), value is "WRONG". Patched pin: Maglev deopts (status 65), value is "TRUE".

python3 chrome-v8-maglev-literal-stale-Abraxas-Labs.py

or:

cd lab && ./run.sh

Last proven line:

SUCCESS CVE-2026-126910 maglev-literal-stale-load cases=object,array CVE-2026-126910-TYPECONF-WITNESS

Image cve-2026-126910:lab, compose project cve126910, loopback 127.0.0.1:18630. Host is linux-arm64 Docker (Apple Silicon).

FilePurpose
lab/poc.htmlobject and array cases from the V8 regress files
lab/poc.pyheaded-off CFT, Maglev flags, differential WRONG vs TRUE
lab/cve-2026-126910.patchV8 CL 8668740 as mailed

Abraxas Labs [email protected]

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.

    891
  • 2

    ghostlock-app

    yukonga/ghostlock-app

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

    547
  • 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.

    369
  • 4

    redforge

    cybersecurity91/redforge

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

    366
  • 5

    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.

    309
  • 6

    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.

    282

Explore all tools

Browse our collection of tools

View all tools →