← Back to Feed CACHED · 2026-08-07 00:42:51 · CACHE_KEY CVE-2026-39868
CVE-2026-39868 · CWE-20 · Disclosed 2026-06-29

This issue was addressed with improved input validation.

ASSESSED — NOISGATE V0.5
Vendor
Reassessed
Verdict:
Do you agree?
01 · The Real Story

A locked door that crumbles once someone is already inside the house

CVE-2026-39868 is a kernel memory corruption flaw in Apple's LaunchServices framework. A malicious application running on the device can trigger improper input validation that crashes the kernel or corrupts kernel memory — potentially enabling privilege escalation or persistent denial of service. The bug affects a wide swath of Apple platforms: iOS/iPadOS < 26.5.2, macOS Tahoe < 26.5.2, macOS Sequoia < 15.7.8, macOS Sonoma < 14.8.8, tvOS < 26.6, visionOS < 26.6, and watchOS < 26.6. It was credited to elite researchers from STAR Labs, Positive Technologies, and Baidu Security — a roster that strongly signals a weaponizable, Pwn2Own-grade primitive rather than a theoretical crash.

The CISA-ADP assigned CVSS of 9.1 (CRITICAL) with AV:N/AC:L/PR:N/UI:N is misleading. Apple's own description says *"an app may be able to cause unexpected system termination or corrupt kernel memory"* — that is a local exploitation scenario requiring code execution on the target device. The Network attack vector is flatly wrong for a bug that requires a malicious app to be installed and running. A corrected vector closer to AV:L/AC:L/PR:L/UI:R drops the base score into the 6–7 range. The severity is real — kernel corruption is always serious — but the delivery friction and local-only attack surface mean CRITICAL overstates the operational risk for a fleet of managed Apple endpoints.

"Kernel corruption via LaunchServices requires local app execution, not network access. HIGH, not CRITICAL."
02 · The Attack Path

4 steps from start to impact.

STEP 01

Deliver malicious app to target device

The attacker must get a crafted application onto the victim's Apple device. On iOS this requires bypassing App Store review, using an enterprise signing certificate, exploiting a WebKit or Safari bug for initial access, or social-engineering the user into sideloading via AltStore/TestFlight. On macOS, the attacker can use a trojanized DMG, phishing email attachment, or supply-chain-compromised Homebrew/pip package. Gatekeeper, Notarization, and XProtect all stand in the way.
Conditions required:
  • Ability to deliver executable code to the target device
  • User interaction to install/launch (or a prior RCE chain)
Where this breaks in practice:
  • iOS App Store review and sandboxing block most malicious payloads
  • macOS Gatekeeper + Notarization quarantine unsigned/un-notarized binaries
  • Enterprise MDM policies restrict app installation sources
  • Mail gateway / web proxy scanning catches known malicious attachments
Detection/coverage: XProtect signature updates, MDM compliance checks for sideloaded apps, endpoint telemetry from CrowdStrike Falcon / Jamf Protect
STEP 02

Execute malicious app on-device

The planted application must actually run. On iOS, sandboxing constrains what APIs the app can reach. On macOS, the app must survive Gatekeeper's first-launch check and user approval dialogs. TCC (Transparency, Consent, and Control) prompts may also gate access to certain subsystem calls depending on how the LaunchServices trigger is constructed.
Conditions required:
  • App passes or bypasses Gatekeeper / App Store sandbox
  • User grants any required TCC permissions
Where this breaks in practice:
  • macOS Sequoia+ tightened Gatekeeper to require drag-to-Applications for unsigned apps
  • iOS sandbox severely restricts inter-process and kernel-adjacent calls
  • EDR agents on managed Macs will flag suspicious process trees
Detection/coverage: Jamf Protect / CrowdStrike Falcon for macOS process execution anomalies, Unified Logging for LaunchServices errors
STEP 03

Trigger LaunchServices input validation flaw

The running app sends a crafted input to LaunchServices that bypasses validation checks, reaching kernel memory. The exact trigger is undisclosed, but the researcher lineup (STAR Labs, Positive Technologies, Baidu Security) and the 'improved input validation' fix language suggest a structured data parsing flaw — likely a malformed bundle, plist, or launch request that the kernel-side handler processes unsafely.
Conditions required:
  • Malicious app must be executing
  • LaunchServices must be reachable from the app's sandbox context
Where this breaks in practice:
  • No public PoC exists as of August 2026
  • Exact trigger conditions are researcher-held; weaponization requires independent rediscovery or binary diffing the patch
Detection/coverage: Kernel panic logs in /Library/Logs/DiagnosticReports, crash telemetry via Apple Business Manager
STEP 04

Corrupt kernel memory or crash system

Successful exploitation corrupts kernel memory, which could yield a kernel read/write primitive for privilege escalation to root/kernel, or at minimum causes a denial-of-service via kernel panic. A kernel write primitive on Apple Silicon with PAC and PPL is non-trivial to weaponize into a full jailbreak, but the corruption itself is confirmed by Apple's advisory. Impact is Integrity (H) + Availability (H) with no confirmed Confidentiality impact.
Conditions required:
  • Step 3 succeeds
  • Kernel memory layout is predictable enough for reliable corruption (vs. random crash)
Where this breaks in practice:
  • Apple Silicon PAC (Pointer Authentication Codes) makes controlled kernel exploitation harder
  • PPL (Page Protection Layer) and KTRR limit what kernel pages can be modified
  • KASLR adds layout randomization overhead to reliable exploitation
Detection/coverage: Kernel panic frequency monitoring, MDM-reported device restart anomalies
03 · Intelligence Metadata

The supporting signals.

In-the-Wild ExploitationNo confirmed exploitation. Not listed on CISA KEV. No campaigns or threat actor usage reported as of 2026-08-07.
Proof of ConceptNo public PoC available. Credited to Ye Zhang (@VAR10CK) of Baidu Security, Vladislav Shevchenko (Positive Technologies), Billy Jheng Bing Jhong & Pan Zhenpeng (@Peterpan0927) of STAR Labs SG — all known for Pwn2Own-grade kernel work. PoC is likely privately held.
EPSS Score0.00936 (~94th percentile). Indicates low but non-trivial predicted exploitation probability within 30 days.
KEV StatusNot listed on CISA Known Exploited Vulnerabilities catalog as of 2026-08-07.
CVSS VectorCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:H (CISA-ADP). noisgate assessment: AV:N is incorrect — Apple's own advisory describes a local app exploitation scenario. A corrected vector would be closer to AV:L/AC:L/PR:L/UI:R/S:U/C:N/I:H/A:H (~6.3).
Affected VersionsiOS/iPadOS < 26.5.2, macOS Tahoe < 26.5.2, macOS Sequoia < 15.7.8, macOS Sonoma < 14.8.8, tvOS < 26.6, visionOS < 26.6, watchOS < 26.6
Fixed VersionsiOS/iPadOS 26.5.2, macOS Tahoe 26.5.2, macOS Sequoia 15.7.8, macOS Sonoma 14.8.8, tvOS 26.6, visionOS 26.6, watchOS 26.6
Scanning / ExposureNot applicable — this is a local exploitation vuln on endpoints, not a network-exposed service. No Shodan/GreyNoise/Censys signatures relevant.
Disclosure Date2026-06-29 (Apple security update release)
Reporting ResearchersYe Zhang (@VAR10CK) / Baidu Security, Vladislav Shevchenko / Positive Technologies, Billy Jheng Bing Jhong & Pan Zhenpeng (@Peterpan0927) / STAR Labs SG Pte. Ltd.
04 · The Call

noisgate verdict.

Final Verdict
DOWNGRADED to HIGH (7.2/10)

The single most decisive factor driving this downgrade from CRITICAL to HIGH is the misattributed attack vector: Apple's own advisory describes local app execution (AV:L), not unauthenticated network exploitation (AV:N) as CISA-ADP scored it — collapsing the reachable population from every-network-host to only devices where the attacker has already achieved code delivery. The kernel memory corruption outcome on Apple endpoints is serious enough to hold the floor at HIGH, but the delivery friction and absence of a public PoC or ITW exploitation prevent CRITICAL.

HIGH Affected version ranges and patch availability
HIGH Attack vector correction (local, not network)
MEDIUM Exploitation complexity and weaponization potential
LOW Whether the corruption yields a reliable kernel r/w primitive vs. crash-only

Why this verdict

  • CVSS vector is wrong at the source. CISA-ADP scored AV:N but Apple's description explicitly says *"an app may be able to"* — that is local code execution, not network-reachable. Correcting to AV:L/PR:L/UI:R alone drops the base from 9.1 to ~6.3, a full severity tier.
  • Delivery friction is substantial. Getting a malicious app onto a managed Apple device requires bypassing Gatekeeper + Notarization (macOS) or App Store review + sandbox (iOS). Enterprise MDM profiles further restrict installation sources. This is not a fire-and-forget network exploit.
  • No public PoC and no ITW exploitation. Despite disclosure 5+ weeks ago, no proof-of-concept code has surfaced publicly, and the bug is not KEV-listed. The elite researcher credits (STAR Labs, Positive Technologies, Baidu) suggest the primitive is real but privately held.
  • Kernel corruption is genuinely dangerous when achieved. A kernel write primitive on macOS could bypass SIP, TCC, and all userland sandboxing. This prevents a downgrade below HIGH despite the delivery friction.
  • Role multiplier: Apple endpoints occupy roles (a) workstations / developer machines and (b) mobile devices for executives and knowledge workers. They are NOT typically identity infrastructure, hypervisors, or supply-chain servers. However, macOS developer workstations with CI/CD access or signing certificates represent a *moderate* high-value role (~5–10% of enterprise Mac fleet). A kernel compromise on a developer Mac with Xcode + signing identity could enable supply-chain pivoting. This keeps the floor at HIGH but does not elevate to CRITICAL because the chain requires prior local code execution AND the high-value role is a minority of the installed base.

Why not higher?

CRITICAL would require either confirmed ITW exploitation, a true network-reachable attack vector, or the affected component being canonically infrastructure (hypervisor, IdP, DC). None apply here. The bug requires local app execution, has no public PoC, is not KEV-listed, and Apple endpoints are overwhelmingly workstation-role devices. The delivery prerequisites impose real friction that compounds against mass exploitation.

Why not lower?

MEDIUM would understate the risk of kernel memory corruption reported by three top-tier offensive research teams. If weaponized, the primitive bypasses every userland security boundary on Apple platforms (SIP, TCC, sandbox). The broad affected surface (iOS + macOS + tvOS + watchOS + visionOS) and the caliber of the researchers who independently found it signal this is not a theoretical crash — it is a serious kernel primitive behind a delivery gate.

05 · Compensating Control

What to do — in priority order.

  1. Enforce MDM app installation restrictions — Configure your MDM (Jamf Pro, Kandji, Mosyle, Intune) to block sideloaded apps and restrict installation sources to the App Store and your internal catalog. This eliminates the primary delivery vector. Deploy within 30 days per the noisgate mitigation SLA for HIGH.
  2. Ensure Gatekeeper and XProtect are enforced fleet-wide — Verify via MDM compliance policy that Gatekeeper is set to 'App Store and identified developers' (minimum) and XProtect definitions are current. spctl --status should return assessments enabled on every Mac. This blocks unsigned/un-notarized payloads.
  3. Push OS updates via MDM enforced deadline — Use MDM-enforced software update policies to push macOS 15.7.8 / 14.8.8 / Tahoe 26.5.2 and iOS/iPadOS 26.5.2 with a 30-day compliance deadline. Devices not updated by deadline should be quarantined from corporate resources via conditional access.
  4. Monitor for kernel panic anomalies — Set up alerting on kernel panic frequency per device via Apple Business Manager telemetry or endpoint agent crash reporting. A spike in kernel panics on a specific device could indicate exploitation attempts.
What doesn't work
  • Network-level controls (firewall, IDS/IPS, WAF) — despite the CVSS AV:N label, this is a local exploitation bug. Network filtering cannot intercept the malicious app-to-kernel interaction.
  • Patching Safari or WebKit alone — the vulnerability is in LaunchServices, not the browser. A Safari update does not remediate this CVE.
  • Disabling LaunchServices — it is a core OS framework required for app launching on all Apple platforms. It cannot be disabled or restricted without breaking the OS.
06 · Verification

Crowdsourced verification payload.

Run this script on each macOS target host (or push via MDM script payload). No special privileges required — it reads sw_vers output. Example: bash check_cve_2026_39868.sh

noisgate-verify.sh
BASHREAD-ONLYSAFE
#!/bin/bash
# check_cve_2026_39868.sh
# Checks whether macOS is patched for CVE-2026-39868 (LaunchServices kernel memory corruption)
# Outputs: VULNERABLE / PATCHED / UNKNOWN
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN

PRODUCT=$(sw_vers -productName 2>/dev/null)
VERSION=$(sw_vers -productVersion 2>/dev/null)

if [ -z "$VERSION" ]; then
  echo "UNKNOWN - unable to determine macOS version"
  exit 2
fi

# Compare version strings
version_gte() {
  # Returns 0 if $1 >= $2
  printf '%s\n%s' "$2" "$1" | sort -V -C
}

MAJOR=$(echo "$VERSION" | cut -d. -f1)

case "$MAJOR" in
  26)
    # macOS Tahoe - fixed in 26.5.2
    if version_gte "$VERSION" "26.5.2"; then
      echo "PATCHED - macOS Tahoe $VERSION >= 26.5.2"
      exit 0
    else
      echo "VULNERABLE - macOS Tahoe $VERSION < 26.5.2 (CVE-2026-39868)"
      exit 1
    fi
    ;;
  15)
    # macOS Sequoia - fixed in 15.7.8
    if version_gte "$VERSION" "15.7.8"; then
      echo "PATCHED - macOS Sequoia $VERSION >= 15.7.8"
      exit 0
    else
      echo "VULNERABLE - macOS Sequoia $VERSION < 15.7.8 (CVE-2026-39868)"
      exit 1
    fi
    ;;
  14)
    # macOS Sonoma - fixed in 14.8.8
    if version_gte "$VERSION" "14.8.8"; then
      echo "PATCHED - macOS Sonoma $VERSION >= 14.8.8"
      exit 0
    else
      echo "VULNERABLE - macOS Sonoma $VERSION < 14.8.8 (CVE-2026-39868)"
      exit 1
    fi
    ;;
  *)
    # macOS 13 or earlier - likely vulnerable but EOL
    echo "UNKNOWN - macOS $VERSION ($MAJOR.x) is not in the patched version matrix. If EOL, assume VULNERABLE."
    exit 2
    ;;
esac
07 · Bottom Line

If you remember one thing.

TL;DR
Patch your Apple fleet within the standard HIGH window, not the vendor's CRITICAL urgency. The CISA-ADP CVSS of 9.1 is inflated by a wrong attack vector — this is local app exploitation, not network-reachable. That said, kernel memory corruption found by three Pwn2Own-tier teams is not something you ignore. Per the noisgate mitigation SLA, enforce MDM app-source restrictions and Gatekeeper compliance within 30 days as your compensating control. Per the noisgate remediation SLA, push macOS 15.7.8 / 14.8.8 / Tahoe 26.5.2 and iOS/iPadOS 26.5.2 to all managed devices within 180 days — ideally faster for developer workstations with signing certificates or CI/CD access, where the blast radius of kernel compromise is supply-chain-adjacent. There is no active exploitation or KEV listing, so this does not warrant emergency out-of-cycle patching.

Sources

  1. NVD - CVE-2026-39868
  2. MITRE CVE Record
  3. Apple Security Update - macOS Sequoia 15.7.8
  4. Apple Security Update - macOS Sonoma 14.8.8
  5. Apple Security Update - iOS/iPadOS 26.5.2
  6. ZDI - June 2026 Apple Security Update Review
  7. Aviatrix Threat Research - June 2026 Apple Updates
Peer Review

What defenders are saying.

Submit a review attribution: handle + country only
0 flags selected · stored anonymously
Validation Results

Crowdsourced verification outputs.

Results submitted by users who ran the verification payload against their environment.