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-2026-112611-cg-shading — CVE-2026-112611/112610: CoreGraphics PDF out-of-bounds writes (fixed in iOS 27.2) — fix-diff analysis, on-device proof, methodology, and PoC generator scripts. No prebuilt PoCs, no crash reports. | Kitploit
Tools/GitHubGitHub/ping-2o/cve-2026-112611-cg-shading
iOS SecurityVulnerability AnalysisExploitationReverse EngineeringFuzzingMobile SecurityBinary AnalysisPapers & Research
Share
GitHub
ping-2o/cve-2026-112611-cg-shading

CVE-2026-112611-cg-shading

CVE-2026-112611/112610: CoreGraphics PDF out-of-bounds writes (fixed in iOS 27.2) — fix-diff analysis, on-device proof, methodology, and PoC generator scripts. No prebuilt PoCs, no crash reports.

View Repository
23725 days agoReviewed by Kitploit

CVE-2026-112611 / CVE-2026-112610 — CoreGraphics PDF out-of-bounds writes (iOS 27.2 fix-diff analysis, with on-device proof)

This repository documents two out-of-bounds write vulnerabilities in CoreGraphics' PDF stack, both fixed in iOS 27.2 (24B115) and credited in Apple's security advisory HT149230 (CVE-2026-112611, CVE-2026-112610):

  • The analysis is an independent reconstruction from Apple's shipped binaries — a function-level diff of the 27.0 (24A435) vs 27.2 (24B115) CoreGraphics dylibs extracted from the two dyld shared caches, verified down to instructions on both sides.
  • The primary site was then proven on an owned, unpatched 27.0 device: a 7-line PDF crashes Preview deterministically, and the crash fingerprint proves the write carries attacker-chosen values (the dereferenced wild pointer is the value the PDF selected).
  • No crash reports, device identifiers, or personal data are included — raw .ips files are intentionally withheld; the receipts in analysis/on-device-proof.md are the technical facts only.

Both CVEs are already fixed and publicly credited by Apple — to Anna Groza (@pwnlambda) of Calif.io and others (CVE-2026-112611) and Nik Tsytsarkin (CVE-2026-112610). This repo adds the missing public artifact: where exactly the bugs were, how the fix closes them, and a reproducible crash oracle (as a generator script — no pre-built PoC files).

The two sites at a glance

Site 1 — Type 5 (lattice-form Gouraud) shadingSite 2 — CMap /UseCMap recursion
CVECVE-2026-112611 (probable; see note)CVE-2026-112610 (probable; see note)
ClassOOB write of attacker-chosen doubles, stack VLA overrunOOB write up the stack via unclamped codespace length
Trigger/Sh1 sh content-stream operator — fires on any page render (preview/thumbnail), no fonts, no text, no user interactionType-0 (CID) text rendering/extraction with an embedded CMap
Pre-fix_cg_shading_type5_draw runs CGFunctionEvaluate per mesh vertex: R function outputs (R = the /Function's /Range dimension) written into a record with space for nc colorspace components_cmap_convert recursion passes the outer codespace length L (accepted up to 255) into out-slots sized for exactly one result (2 B code / 8 B len)
Fix (27.2)Parse-time gate range_dim == nc && nc <= 4 + CG-SEC-187882049 log; draw code byte-identical → gate-only fixmin(L,4) clamp + out-slot re-layout + CG-SEC-187882136 log
On-device resultProven: deterministic EXC_BAD_ACCESS, fault 0x3ff000000000001c = the bit pattern of the PDF-chosen double 1.0 (0x3ff0000000000000) + 0x1c — attacker value controlVehicle refuted for embedded CMaps (three independent reachability blockers, see analysis) — the credited researcher's actual vehicle is presumably different

Note on CVE assignment. Apple credits two OOB-write fixes under CoreGraphics and the 27.2 diff contains exactly two new bounds checks — a 1:1 correspondence. Which check is which CVE is not statically determinable; the assignment above is our ranking (the shading site is the stronger 1-click primitive), not evidence.

Repository layout

README.md                              this file
analysis/CVE-2026-112611_type5_shading.md    full site-1 writeup (both-side disasm receipts)
analysis/CVE-2026-112610_cmap_recursion.md   full site-2 writeup + reachability analysis
analysis/on-device-proof.md                  the device run: receipts, fingerprint decode, verdicts
analysis/methodology.md                     the fix-diff method, reusable
poc/make_poc_shading.py                      generator: builds the shading oracle + control PDFs
poc/make_poc_cmap.py                        generator: builds the CMap oracle + control PDFs
tools/ips_parse.py                          generic Apple .ips crash-report parser

Building the PoCs

No crafted PDFs are committed. Generate them locally (needs nothing but Python 3):

python3 poc/make_poc_shading.py   # -> poc/crash_shading_type5.pdf + poc/control_shading_type5.pdf
python3 poc/make_poc_cmap.py      # -> poc/oracle_cmap*.pdf + poc/control_cmap.pdf

The shading oracle is a 7-object PDF (~2 KB) whose only unusual content is:

  • a /ShadingType 5 dictionary with /ColorSpace /DeviceGray (nc = 1) and
  • a /FunctionType 2 function whose /Range declares 136 output dimensions instead of 1,
  • with /C0 and /C1 all 1.0 — so every out-of-bounds double written is 0x3FF0000000000000, and any wild pointer built from those spill slots dereferences a recognizable, chosen address.

The page content stream is a single operator: /Sh1 sh.

Verifying (differential oracle)

BuildPreview the crash PDFExpected receipt
iOS 27.0 / 27.0.x (≤ 24A446, pre-fix)crashes the preview/render processEXC_BAD_ACCESS KERN_INVALID_ADDRESS at 0x3ff000000000001c (deterministic; see analysis/on-device-proof.md)
iOS 27.2 (24B115, fixed)renders cleanlyunified log: CG-SEC-187882049: invalid function for type 5 shading (range dimension does not match color space).

To check the fixed-side log line on a 27.2 device:

log show --last 15m --predicate 'processImagePath CONTAINS[c] "quicklook" OR processImagePath CONTAINS[c] "preview"' --info --debug | grep CG-SEC

Crash reports (.ips) can be inspected with tools/ips_parse.py; on the pre-fix build the faulting frame symbolicates to cg_shading_type5_draw, called from op_sh → CGPDFScannerScan → the page-background render path.

Methodology (summary)

The full method is in analysis/methodology.md. In short:

  1. Extract the same dylib (CoreGraphics) from both builds' dyld shared caches (ipsw dyld extract).
  2. Strings diff first — new assert/log strings mark fix sites; here two new CG-SEC-187882049 / CG-SEC-187882136 strings localize both fixes immediately.
  3. Function-level diff with relocation-normalized instruction hashing; classify every PDF-stack change; verify nothing else is security-relevant.
  4. Both-side disassembly of each site; reconstruct the pre-fix dataflow instruction by instruction; label everything PROVEN vs SHAPE-MATCHED.
  5. Build a crash-only oracle with a value fingerprint (chosen IEEE-754 constants whose bit patterns the crash must exhibit) + a benign control; verify on device; publish negatives too.

Responsible use

  • The bugs are already fixed by Apple and publicly credited; this repository contains analysis and a crash reproducer only — no exploitation logic, no payloads, no weaponization.
  • Test only on devices you own. The oracle crashes a preview process; it does not execute code.
  • Tested platforms: iPhone 17,x on iOS 27.0 RC (24A435). Fixed behavior verified statically against the 27.2 RC (24B115) binary.

References

  • Apple: About the security content of iOS 27.2 and iPadOS 27.2 (CVE-2026-112611, CVE-2026-112610, released 2026-10-26)
  • Apple credits: Anna Groza (@pwnlambda) of Calif.io et al. (CVE-2026-112611); Nik Tsytsarkin (CVE-2026-112610)
  • Build IDs compared: CoreGraphics iOS 27.0 (24A435, LC_SOURCE_VERSION 2050.0) vs iOS 27.2 RC (24B115, LC_SOURCE_VERSION 2050.1.6)
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.

    1.0k
  • 2

    ghostlock-app

    yukonga/ghostlock-app

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

    627
  • 3

    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.

    416
  • 4

    jumpserver

    jumpserver/jumpserver

    JumpServer is an open-source Privileged Access Management (PAM) platform that provides DevOps and IT teams with on-demand and secure access to SSH, RDP, Kubernetes, Database and RemoteApp endpoints through a web browser.

    379
  • 5

    ARTEX

    baarai/artex

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

    330
  • 6

    Aliens_eye

    arxhr007/aliens_eye

    AI powered OSINT engine to discover and map 7,700+ social media identities across the web

    288

Explore all tools

Browse our collection of tools

View all tools →