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
Tools/GitHubGitHub/mchinamd57764/cve-2026-112706-coremedia-fd-leak
Privilege EscalationDynamic Analysis (Sandboxing)iOS SecurityVulnerability AnalysisExploitationMobile App PentestingInformation GatheringMobile Security
Share
Papers & Research
GitHubmchinamd57764/cve-2026-112706-coremedia-fd-leak

CVE-2026-112706-CoreMedia-fd-leak

Independent discovery & coordinated disclosure: CoreMedia bytestream XPC (mediaplaybackd) vended live file descriptors for protected paths to any sandboxed app on iOS 26.5–27.0.1. Fixed in iOS 27.2 as CVE-2026-112706. PoC + evidence.

View Repository
27824 days agoReviewed by Kitploit

CVE-2026-112706: CoreMedia bytestream XPC leaks live file descriptors to any sandboxed app

Status: FIXED in iOS 27.2 / iPadOS 27.2 (and tvOS/visionOS/watchOS 27.2), released 2026-10-27. Affected (verified): iOS 26.5 (23F77), iOS 27.0 (24A437), iOS 27.0.1 (24A446).

This repository documents an independent discovery and coordinated disclosure of this issue.

Timeline

DateEvent
2026-10-15Issue independently discovered and reported to Apple Product Security (report OE11044060567043).
2026-10-27Apple ships the fix in iOS/iPadOS/tvOS/visionOS/watchOS 27.2 as CVE-2026-112706 ("An app may be able to access sensitive user data" — CoreMedia, "a permissions issue was addressed with additional restrictions").
2026-10-27Advisory credits the earlier report: Adriatik Raci, Rodolphe BRUNETTI (@eisw0lf) of Lupus Nova, Chris Bailey – Short Circuit. My report was confirmed as a duplicate of a previously known issue; per Apple policy the duplicate received no bounty and no advisory credit.
2026-10-28Post-fix publication of this writeup and PoC for the record.

I am publishing this because the advisory does not reflect my independent report, and the material here may be useful to researchers studying XPC/sandbox design flaws on Apple platforms. Nothing in this repository is zero-day: the issue is patched on all current releases. Do not test the PoC against devices you do not own.

Summary

Every iOS device exposes the launchd Mach service com.apple.coremedia.mediaplaybackd.bytestream.xpc. A sandboxed third-party app — no entitlements, no permission prompts, no usage-description keys — can connect to it directly and obtain live file descriptors for files outside its sandbox, opened by the mediaplaybackd daemon with its own credentials.

The chain:

  1. Unauthenticated origin handshake. The FigXPC memory-origin handshake (.rp1 → .rp2 → .org) completes with status 0 for an unprivileged client.
  2. Curl create with a client-controlled path. The create message carries a "CreationURL" dictionary with an absolute file:// path chosen by the client. The daemon opens that path itself, never validating it against the requesting client's sandbox container.
  3. Live fd in the reply. The reply's ByteStreamDetails dictionary contains a live FileDescriptor for the opened file. The app dups it and pread()s it — descriptor-based I/O bypasses the client sandbox's path checks, so the full content of the protected file lands in the app container.
  4. Bonus primitives. The same request shape yields directory fds (fdopendir/readdir enumeration of protected directories) and a file-existence oracle (a real-but-forbidden path returns -12203, a nonexistent path -17913).

Control evidence: in the same run, the app's own open() of the identical path fails with errno 1 (EPERM). The daemon-vended fd returns every byte.

Incomplete mitigation in iOS 27.0: on iOS 26.5 the auxiliary opcodes read / fdsc / getL were also open; iOS 27.0 sealed those (-16158) but left the primary fd-vending path inside the Curl create reply fully functional — verified end-to-end on 27.0 and 27.0.1.

Readable domain (mapped on both versions, 27 targets probed): exactly the media-library directory /private/var/mobile/Media/iTunes_Control/iTunes/ (database trio, PlayCounts.plist, VoiceMemos.plist, per-user ContentTaste plists, IC-Info device files). User photos, messages, contacts, the TCC permission database, keychain, Safari data, Health, Wallet passes and other apps' containers were all refused with -12203. The write path was denied on both versions (-12878), so the primitive is read/enumerate/oracle, not write.

Evidence

All logs are from clean, stock (non-jailbroken) test devices. Unrelated third-party tool residue in directory listings is replaced with [redacted].

FileDevice / OSWhat it shows
evidence/wave4_clean_16pro_result.txtiOS 26.5 (23F77)EPERM control, fd steal 749,568 + 2,175,392 B, read data channel, dir list (18), oracle
evidence/wave4_ios270_result.txtiOS 27.0 (24A437)Curl OK, fd in ByteStreamDetails, steals 835,584 + 1,730,432 + 32,768 B, aux opcodes sealed -16158, dir list (13), oracle
evidence/wave4_ios2701_result.txtiOS 27.0.1 (24A446)Re-run after update: Curl/fd path still live (835,584 + 1,862,272 B) — not fixed in 27.0.1
evidence/wave5_domain_map.txtboth27-target readable-domain map; only iTunes_Control/iTunes/ opens
evidence/wave4_ab_matrix.mdbothOne-page 26.5 vs 27.0 channel matrix

The stolen database copies themselves are not published (test-device data); the PoC regenerates them on any affected device.

PoC

poc/ contains a minimal Xcode project (MConfPoC). To run:

  1. Open in Xcode, set your development team (the project intentionally has YOUR_TEAM_ID placeholders).
  2. Build & run on a device running an affected iOS version (≤ 27.0.1).
  3. No interaction is needed: the chain fires on launch and writes Documents/wave4_clean_result.txt plus the stolen .bin copies.
  4. Pull the results:
    xcrun devicectl device copy from --device <UDID> \
      --source Documents/wave4_clean_result.txt --destination . \
      --domain-type appDataContainer --domain-identifier <bundle-id>
    

Core chain (~390 lines of Swift): dlopen CoreMedia + dlsym the FigXPC helpers → xpc_connection_create_mach_service → .rp1/.rp2/.org handshake → FigXPCCreateBasicMessage('Curl') + FigXPCMessageSetCFURL("CreationURL") → reply ByteStreamDetails.FileDescriptor → dup/fstat/pread.

Suggested fix (as reported to Apple)

  1. On every Curl create, verify the requesting process (audit token / sandbox ownership) and reject paths the client itself cannot open; never open client-supplied absolute paths with daemon credentials.
  2. Apply the same gating to the FileDescriptor embedded in the create reply as was applied to the read/fdsc opcodes in iOS 27.0.
  3. Extend the check to directory opens (enumeration primitive).
  4. Return a uniform error for "denied because exists" vs "does not exist" to remove the existence oracle.

References

  • Apple — About the security content of iOS 27.2 and iPadOS 27.2 (CVE-2026-112706)
  • Apple — About the security content of tvOS 27.2 / visionOS 27.2 / watchOS 27.2 (same CVE)

Disclaimer

Provided for educational and defensive research purposes only. The vulnerability is patched; running the PoC on current iOS will simply fail. The author is not responsible for misuse.

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)

    647
  • 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.

    421
  • 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.

    389
  • 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.

    333
  • 6

    SafeLine

    chaitin/safeline

    Self-hosted WAF and reverse proxy that filters malicious HTTP traffic, blocks SQL injection, XSS, and bot attacks, with rate limiting and dynamic HTML/JS encryption.

    331

Explore all tools

Browse our collection of tools

View all tools →