Skip to content

Security Sense

AI Threat Intelligence

  • About
  • Blog
  • Privacy Policy

Tag: macos

Apple to Tighten macOS Full Disk Access Controls Over AI Agent Misuse

Posted on October 5, 2026 by Kyle

Overview

Apple is addressing a significant security risk on macOS related to the Full Disk Access (FDA) permission. Certain AI agents are leveraging FDA in ways that grant them overly broad access to user data, including personal files, emails, messages, and browsing history, often without the user’s full understanding. This initiative aims to close a potential privacy and security gap that could lead to widespread data exposure.

Technical Analysis

  • Vulnerability: Misuse of the macOS Full Disk Access (FDA) permission by applications, specifically AI agents.
  • Mechanism: FDA is a legitimate macOS security feature designed to allow applications (e.g., backup software, antivirus) to access protected user data directories.
  • Exploitation: Developers of some AI agents are requesting and receiving FDA, which then grants their applications unrestricted access to sensitive user data across the system. This access includes /Users/<username>/Library/Mail, /Users/<username>/Library/Messages, Safari history, and other personal files.
  • Prerequisites: The user must grant FDA to the AI agent application. The core issue is that the scope of this permission is often not fully understood by the user, and the application may not explicitly inform the user about the full extent of data it will access.
  • Impact: Unauthorized access to sensitive user data, potential for data exfiltration, privacy violations, and compromise of system integrity if the AI agent is malicious or compromised.

Detection

  • Monitor macOS unified logs for TCC (Transparency, Consent, and Control) database changes related to FDA grants, if available via EDR or audit logs.
  • Review tccutil output or TCC.db for applications with kTCCServiceSystemPolicyAllFiles permission.
  • Behavioral analysis of applications granted FDA: look for unusual file access patterns, excessive reads from sensitive directories (e.g., mail, messages, browser history), or network connections to untrusted destinations.
  • Endpoint Detection and Response (EDR) solutions can track process activity and file access by applications with FDA.

Sigma Detection Rules

⚠️ AI-generated detection rules. These are experimental starting points. Review field names, EventIDs, and logic against your environment’s schema before deploying. Tune to reduce false positives.

macOS Application Accessing Sensitive User Data Directories

title: macOS Application Accessing Sensitive User Data Directories
id: 5a8b3c2d-e1f2-4a3b-9c0d-1e2f3a4b5c6d
status: experimental
description: Detects applications accessing sensitive user data directories on macOS, potentially indicating misuse of Full Disk Access by AI agents or other applications.
logsource:
  product: macos
  service: file_access
detection:
  selection_mail:
    TargetFilename|contains: '/Library/Mail/'
  selection_messages:
    TargetFilename|contains: '/Library/Messages/'
  selection_safari_history:
    TargetFilename|contains: '/Library/Safari/History.db'
  condition: 1 of selection_*
level: high

Mitigations

  1. Review FDA Grants: Regularly audit applications listed under System Settings > Privacy & Security > Full Disk Access. Revoke FDA for any application not explicitly requiring it or for those whose purpose does not justify such broad access.
  2. Exercise Caution with AI Agents: Be highly selective when installing and granting permissions to AI agent applications, especially those requesting FDA. Understand the data access implications before approval.
  3. Monitor for Apple Updates: Apply macOS security updates promptly as Apple implements tighter controls around FDA and AI agent data access.
  4. Least Privilege: Adhere to the principle of least privilege for all applications. Only grant necessary permissions.
  5. Data Minimization: Limit the amount of sensitive data stored locally where possible, especially if using applications that require broad system access.

References

  • https://thehackernews.com/2026/10/apple-plans-tighter-macos-full-disk.html

Indicators of Compromise

No public IOCs available at time of writing.

MITRE ATT&CK

  • T1083 — File and Directory Discovery
🤖 AI Attribution
Generated by gemini-2.5-flash ·
1,533 input / 1,069 output tokens ·
Reviewed and approved by a human analyst before publication
Posted in vulnerabilityTagged ai, critical, data-exposure, macos, privacy, vulnerabilityLeave a comment

Apple Emergency Patch for iOS 26, macOS 26, macOS 15 (CVE-2026-86950)

Posted on September 30, 2026 by Kyle

Overview

Apple has issued urgent security updates for older versions of its operating systems, specifically iOS 26, macOS 26, and macOS 15. These patches address CVE-2026-86950, a vulnerability confirmed to be under active exploitation. The immediate patching requirement underscores the critical nature of this flaw for users running the affected legacy branches.

Technical Analysis

Details regarding the specific nature of CVE-2026-86950 have not been publicly disclosed by Apple at the time of writing, beyond its active exploitation status. The vulnerability affects:
* iOS 26
* macOS 26
* macOS 15

Notably, the current generation iOS 27 and macOS 27 operating systems are explicitly stated as not being affected by this vulnerability. The attack vector and prerequisites for exploitation are currently unknown, but the emergency patch suggests a high-impact flaw, likely leading to arbitrary code execution or privilege escalation.

Detection

Specific detection methods for CVE-2026-86950 are challenging without public technical details of the exploit. However, defenders should focus on post-exploitation behaviors on affected macOS systems:
* Unusual Process Spawns: Monitor for unexpected child processes originating from common user applications (e.g., browsers, mail clients) or system services that typically do not execute shell commands or create new executables.
* Privilege Escalation Attempts: Look for attempts to modify system files, create new privileged users, or execute commands with elevated permissions following suspicious activity.
* Network Connections: Observe outbound network connections from unusual processes or to suspicious external IP addresses/domains.
* File System Monitoring: Monitor for creation of suspicious files in temporary directories, user library folders, or system directories, especially those with executable permissions.

Sigma Detection Rules

⚠️ AI-generated detection rules. These are experimental starting points. Review field names, EventIDs, and logic against your environment’s schema before deploying. Tune to reduce false positives.

macOS Suspicious Child Process from Common Applications

title: macOS Suspicious Child Process from Common Applications
id: 76f8e2c3-1a2b-4d5e-8f9a-0b1c2d3e4f5a
status: experimental
description: Detects suspicious child processes (e.g., shells, scripting interpreters) spawned by common user applications, which could indicate post-exploitation activity on macOS. This is a generic rule for client-side exploits leading to code execution.
logsource:
  product: macos
  service: es_process_events
detection:
  selection_parent:
    ParentImage|contains:
      - '/Applications/Safari.app/'
      - '/Applications/Mail.app/'
      - '/Applications/Messages.app/'
      - '/Applications/Firefox.app/'
      - '/Applications/Google Chrome.app/'
      - '/Applications/Microsoft Word.app/'
      - '/Applications/Microsoft Excel.app/'
      - '/Applications/Microsoft PowerPoint.app/'
  selection_child:
    Image|endswith:
      - '/bin/sh'
      - '/bin/bash'
      - '/bin/zsh'
      - '/usr/bin/python'
      - '/usr/bin/perl'
      - '/usr/bin/ruby'
      - '/usr/bin/osascript'
      - '/usr/bin/curl'
      - '/usr/bin/wget'
  condition: all of selection_*
level: high

Mitigations

  1. Apply Patches Immediately: Update all affected iOS 26, macOS 26, and macOS 15 devices to the latest security patches provided by Apple. This is the most critical and effective mitigation.
  2. Isolate Legacy Systems: If immediate patching is not feasible, isolate affected legacy systems from critical networks and sensitive data until updates can be applied.
  3. Endpoint Detection and Response (EDR): Ensure EDR solutions are deployed and configured to monitor for suspicious process activity, file modifications, and network connections on macOS endpoints.
  4. User Awareness: Educate users on phishing and social engineering tactics, as initial access for such exploits often relies on user interaction.

References

  • https://isc.sans.edu/diary/rss/33376

Indicators of Compromise

No public IOCs available at time of writing.

🤖 AI Attribution
Generated by gemini-2.5-flash ·
1,599 input / 1,186 output tokens ·
Reviewed and approved by a human analyst before publication
Posted in cveTagged apple, critical, cve, ios, macos, patch, vulnerabilityLeave a comment

CVE-2026-86950: Apple CoreGraphics Out-of-Bounds Write Leading to RCE

Posted on September 30, 2026 by Kyle

Overview

CVE-2026-86950 is an out-of-bounds write vulnerability in Apple’s CoreGraphics framework. This critical flaw affects iOS, iPadOS, and macOS, allowing arbitrary code execution when processing a maliciously crafted file. Apple has confirmed reports of this vulnerability being exploited in sophisticated, targeted attacks, underscoring the urgency for immediate patching.

Technical Analysis

The vulnerability, tracked as CVE-2026-86950 (CWE-787), stems from insufficient bounds checking within the CoreGraphics component. When a user processes a specially crafted file, this flaw can be triggered, leading to an out-of-bounds write operation. This memory corruption can then be leveraged to achieve arbitrary code execution in the context of the vulnerable application.

  • Vulnerability Type: Out-of-bounds Write (CWE-787)
  • Affected Products:
    • Apple iOS versions prior to 26.7.1
    • Apple iPadOS versions prior to 26.7.1
    • Apple macOS versions prior to 15.8.1
    • Apple macOS versions 26.0 through 26.7.0
  • Attack Vector: Network (AV:N), User Interaction Required (UI:R). Exploitation occurs when a user processes a maliciously crafted file. This could be via email attachments, malicious websites, or messaging applications.
  • Impact: Arbitrary Code Execution (C:H, I:H, A:H). Successful exploitation grants an attacker the ability to execute arbitrary code with the privileges of the exploited application.
  • CVSS 3.1 Score: 8.8 (High) – CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

Detection

Detecting the initial trigger of this vulnerability (processing a malicious file) can be challenging without specific file hashes or content signatures. Detection efforts should focus on post-exploitation activities, such as unusual process creation or network connections originating from applications that typically handle user-supplied content.

  • Process Monitoring: Monitor for suspicious child processes spawned by applications like Safari, Mail, Messages, Preview, or other image/document viewers. Look for shell processes (sh, bash, zsh), scripting interpreters (python, perl, ruby), or network utilities (curl, wget) originating from these parent processes.
  • Network Activity: Monitor for outbound network connections from typically sandboxed or user-facing applications to unusual external IP addresses or domains.
  • File System Monitoring: Look for creation of suspicious files in temporary directories or user home directories by applications that do not typically write executable content.
  • Endpoint Telemetry: Analyze EDR logs for process execution anomalies, especially those involving CoreGraphics or related libraries.

Sigma Detection Rules

⚠️ AI-generated detection rules. These are experimental starting points. Review field names, EventIDs, and logic against your environment’s schema before deploying. Tune to reduce false positives.

macOS Suspicious Shell Spawn from Apple App

title: macOS Suspicious Shell Spawn from Apple App
id: 8e9f0c1a-2b3d-4e5f-6a7b-8c9d0e1f2a3b
status: experimental
description: Detects shell processes (sh, bash, zsh) spawned by common Apple applications that handle user-supplied content. This could indicate post-exploitation activity following a client-side vulnerability like CVE-2026-86950.
logsource:
  product: macos
  category: process_creation
detection:
  selection_parent:
    ParentImage|contains:
      - '/Applications/Safari.app/'
      - '/Applications/Mail.app/'
      - '/Applications/Messages.app/'
      - '/Applications/Preview.app/'
      - '/Applications/TextEdit.app/'
      - '/Applications/Pages.app/'
      - '/Applications/Numbers.app/'
      - '/Applications/Keynote.app/'
  selection_child:
    Image|endswith:
      - '/bin/sh'
      - '/bin/bash'
      - '/bin/zsh'
  condition: all of selection_*
level: high

Mitigations

  1. Patch Immediately: Apply the latest security updates for all affected Apple devices.
    • Update iOS and iPadOS to version 26.7.1 or later.
    • Update macOS Sequoia to version 15.8.1 or later.
    • Update macOS Tahoe to version 26.7.1 or later.
  2. User Awareness Training: Educate users about the risks of opening unsolicited or suspicious files, attachments, and links, even from seemingly trusted sources.
  3. Endpoint Security: Ensure endpoint detection and response (EDR) solutions are deployed and configured to monitor for anomalous process behavior and network connections.
  4. Application Sandboxing: Leverage built-in application sandboxing features where possible to limit the impact of successful exploitation.
  5. Network Segmentation: Implement network segmentation to restrict outbound communication from user endpoints, limiting potential command-and-control (C2) channels.

References

  • https://support.apple.com/en-us/149226
  • https://support.apple.com/en-us/149228
  • https://support.apple.com/en-us/149229
  • http://seclists.org/fulldisclosure/2026/Sep/89
  • http://seclists.org/fulldisclosure/2026/Sep/90
  • http://seclists.org/fulldisclosure/2026/Sep/91
  • https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-86950

Indicators of Compromise

No public IOCs available at time of writing.

MITRE ATT&CK

  • T1204.002 — Malicious File
  • T1059.004 — Unix Shell
🤖 AI Attribution
Generated by gemini-2.5-flash ·
2,000 input / 1,611 output tokens ·
Reviewed and approved by a human analyst before publication
Posted in vulnerabilityTagged apple, code execution, high, ios, ipados, macos, vulnerabilityLeave a comment
Proudly powered by WordPress | Theme: micro, developed by DevriX.