Skip to content

About

A comprehensive, standalone PowerShell security posture auditor for Windows 11 Enterprise endpoints.

Topics

Resources

Security policy

Stars

6 stars

Watchers

0 watching

Forks

Repository files navigation

EndpointPostureCheck

Comprehensive, standalone, zero-dependency security posture auditors for Windows 11 Enterprise and macOS 15/26 endpoints — one PowerShell, one zsh, no third-party dependencies for either.

This repository contains two independent tools sharing one design philosophy (fault-isolated checks, weighted compliance scoring, a self-contained HTML dashboard, optional JSON export and generated remediation script):

Version 1.1.0 · What's new · Changelog. Both scripts print their version in the console banner, the report header and the JSON export (ToolVersion).

What's new

Version 1.1.0 (2026-11-11). Older versions: CHANGELOG.md.

  • Security fix: a USB, Bluetooth or Wi-Fi name crafted to contain a report template token could add a script to the HTML report. Collected values can no longer open a token.
  • Dashboard: findings grouped under a heading per domain, the sidebar is an index that follows your scrolling and can be collapsed, and a search bar pinned to the top filters the findings.
  • USB and Wi-Fi (both editions): inventories of connected USB and Bluetooth devices and remembered Wi-Fi networks; Windows adds the current connection's security, Mobile Hotspot and removable storage restrictions; macOS adds open or stale known networks, unencrypted external volumes and accessory approval.
  • Evidence everywhere: every finding on both editions names what it read (commands, registry values, preference keys).
  • Windows named pipes: permissions are read without letting the pipe's server impersonate the auditor, and pipes writable by Everyone are listed for review.
  • macOS CIS Intune benchmark: 33 more MDM payload settings, plus Apple Intelligence restrictions.
  • Fixes: the Mobile Hotspot finding on Windows PowerShell 5.1, Wi-Fi checks that passed without data, stricter removable storage scoring, and macOS cleanup of temporary files when a run is interrupted. See the changelog for the full list.

EndpointPostureCheck for Windows

A comprehensive, standalone PowerShell security posture auditor for Windows 11 Enterprise endpoints.

EndpointPostureCheck runs 180 read-only security controls across 36 hardening domains — from firmware and virtualization-based security through privilege-escalation vectors, AppLocker bypass paths, ransomware resilience, and modern cloud-identity token security — and renders the results as a single-file, interactive HTML dashboard. Every finding is mapped to CIS Controls v8, NIST SP 800-53 Rev.5, MITRE ATT&CK, ISO/IEC 27001:2022, the NIS2 Directive, and NSA/CISA guidance, and the tool can generate a companion remediation script for everything it finds wrong.

It is designed to run standalone with zero external dependencies — no modules to install, no internet access required, just PowerShell and the Windows APIs already on the box.


Contents


Highlights

  • 180 checks, zero dependencies. Everything runs on native PowerShell, CIM/WMI, and built-in Windows command-line tools (bcdedit, reagentc, vssadmin, netsh, auditpol, wevtutil, sc, secedit). Nothing to install.
  • Fault-tolerant by design. Every single check runs inside its own try/catch boundary. A missing subsystem, an unsupported SKU, or an access-denied condition is logged as a WARN/INFO finding — it never aborts the run.
  • Executive HTML dashboard. A self-contained, single-file dashboard in the Låda Glass family design (frosted glass, dark theme with a light theme that follows the system, and a theme switch). It shows a compliance score ring, risk badge, KPI cards that filter by status, a sidebar listing every domain with its FAIL/WARN count, a System Overview panel (hostname, domain/Entra ID join state, uptime, IP, disk space, RAM/CPU, installed app count, BitLocker status), and a searchable, filterable findings table. Long item lists open in a popover. On a phone the table becomes stacked cards, and it has print styles. The design system's CSS and icons are inlined, so the report makes no network requests.
  • Weighted compliance scoring. Critical/High severity failures are penalized more heavily than Medium/Low ones; the result is a single 0–100% score and a LOW/MEDIUM/HIGH/CRITICAL risk badge.
  • Multi-framework mapping on every finding. CIS Controls v8, NIST SP 800-53 Rev.5, MITRE ATT&CK technique IDs, ISO/IEC 27001:2022 Annex A, NIS2 Article 21(2), and NSA/CISA guidance.
  • Auto-generated remediation script. -GenerateRemediationScript produces a companion Remediate-EndpointPostureCheck.ps1 with a commented, ShouldProcess-gated block for every FAIL finding — real executable commands where it's safe to automate, explicit manual-review guidance where it isn't.
  • JSON export for feeding results into a SIEM, ticketing system, or a central compliance dashboard.

Requirements

Requirement Detail
OS Windows 11 (Enterprise/Education recommended for full AppLocker/WDAC coverage; most checks also run on Pro)
PowerShell 5.1 (Windows PowerShell) or 7.x
Privileges Local Administrator (the script enforces this via #Requires -RunAsAdministrator and a redundant runtime check)
Modules None — no Install-Module required
Network None required — fully offline-capable

Quick Start

# Run a full audit, write the HTML dashboard to the current directory
.\Invoke-EndpointPostureCheck.ps1

# Full audit + JSON export + auto-generated remediation script, written to a specific folder
.\Invoke-EndpointPostureCheck.ps1 -ReportPath C:\Reports -ExportJson -GenerateRemediationScript

The script must be run from an elevated PowerShell session ("Run as Administrator"). If it isn't, it halts immediately with a clear message rather than continuing with a partial, misleading result.


Parameters

Parameter Type Description
-ReportPath string Optional output location for the HTML dashboard. Accepts a directory (default filename is used) or a full file path. Defaults to the current working directory.
-ExportJson switch Exports the complete structured finding set — plus posture/summary metadata — to a JSON file alongside the HTML report.
-GenerateRemediationScript switch Generates Remediate-EndpointPostureCheck.ps1, containing a remediation block for every FAIL finding from the current run.
-BaselinePath string A JSON export (-ExportJson) from an earlier run. Adds two findings: New Privileged & Persistence Items Since Baseline (scored WARN when the autostart inventory, local administrators, local accounts, WMI subscriptions, Defender exclusions, browser extensions or PowerShell profiles gained items) and Posture Changes Since Baseline (informational: controls that got worse or better, and changes in installed applications and inbound firewall rules).
-ExceptionPath string A JSON file of accepted risks (see Accepted risks). Matching FAIL/WARN findings keep their status, are marked Accepted risk and are left out of the score and risk badge until their expiry date. One Accepted Risk Exceptions finding lists what was applied, expired, unmatched or invalid.

Output Files

File Description
EndpointPostureCheck_<ComputerName>_<Timestamp>.html The interactive dashboard. Always generated.
EndpointPostureCheck_<ComputerName>_<Timestamp>.json Full structured export of every finding plus scoring metadata. Generated with -ExportJson.
Remediate-EndpointPostureCheck.ps1 Companion remediation script. Generated with -GenerateRemediationScript. Supports -WhatIf and requires -Force to apply changes.

Understanding the Report

Status — every control resolves to one of:

Status Meaning
PASS The control meets the recommended configuration.
FAIL The control does not meet the recommended configuration and is scored accordingly.
WARN A softer finding — a control that's partially met, context-dependent, or an awareness item worth reviewing rather than an outright failure.
INFO Contextual/inventory information with no pass-fail judgement (e.g. installed application count, current Wi-Fi SSID). Excluded from scoring.

Severity — Critical, High, Medium, Low, or Informational. Severity drives both the scoring weight and the color-coded badge in the dashboard.

Compliance Score — a weighted percentage: PASS earns full credit for a control's severity weight, WARN earns half credit, FAIL earns none. INFO-status findings are excluded entirely. Critical and High severity controls are weighted far more heavily than Low ones, so a handful of critical gaps will visibly move the score even if most controls pass.

Risk Badge:

Badge Condition
CRITICAL Any Critical-severity FAIL present
HIGH 3+ High-severity FAILs, or score below 60%
MEDIUM Score below 80%
LOW Otherwise

Findings are grouped under one heading per domain. The sidebar is an index of those headings: a click scrolls to the domain (it does not filter), and the domain on screen is highlighted as you scroll. A handle at the middle of the sidebar's edge slides it out of the way and back. A search bar stays pinned to the top of the page (press / to jump to it): typing hides every finding that does not match, and domains left empty, and shows how many match. The dashboard lets you filter by status (All / Failures / Warnings / Critical-High, or the KPI cards), filter by domain (the drop-down), and free-text search across control names, categories, measured values, and remediation guidance.

Evidence — where a finding knows where its value came from (the macOS managed-preference checks say whether a configuration profile forces the setting; the Windows policy-source check names the Group Policy and MDM stores), the dashboard shows it under the measured value as Source: ..., and the JSON export carries it as Evidence. Every finding has one on both platforms: checks that do not name their source get what they read (registry values on Windows; commands, preference keys and plist paths on macOS), or else the tools and paths their code queries.

Inventory — below the findings, the dashboard has an Inventory section with sortable, searchable tables: processes (owner, CPU, memory, binary, signature, listeners, hosted services, redacted command line), services, network connections, network exposure (every non-loopback listener, the firewall rules that admit it and the firewall's decision), firewall rules (every rule in the effective policy, inbound and outbound, with program, service, package, ports, addresses, profiles, edge traversal and source store), firewall profiles, named pipes, language runtimes, USB and Bluetooth devices, Wi-Fi profiles and (macOS) external volumes. Tags such as unsigned, writable binary or reachable filter the rows, linked values (a PID, a service, a rule) open the matching rows of another table, a row click shows every field, and CSV downloads the current view. Inventories are never scored; the findings derived from them are. The JSON export carries the same tables under Inventory. Command lines are redacted by pattern, but review the report before sharing it.

Accepted risks

Pass -ExceptionPath (Windows) or --exceptions (macOS) a JSON file like ci/sample-exceptions.json:

{"exceptions": [
  {"control": "Firewall Rules - Risky Allow Rules", "reason": "Lab VLAN, reviewed quarterly",
   "approvedBy": "Security team", "expires": "2027-03-31"}
]}

control must match a control name exactly as the report shows it (case does not matter); reason and expires (YYYY-MM-DD) are required. Accepted findings stay visible with an Accepted risk chip and the reason; the score card shows the score with them included too. An expired entry puts the finding back into the score and marks it Acceptance expired.


Audit Domains

180 checks organized across 36 domains (plus the optional baseline comparison):

# Domain
1 Hardware, Firmware & Platform Integrity (TPM, Secure Boot, Kernel DMA Protection)
2 Virtualization-Based Security & Kernel Isolation (VBS, HVCI, Credential Guard, System Guard, LSA Protection)
3 Volume Encryption & Data-at-Rest Security (BitLocker status, cipher, key protectors, encryption type)
4 Next-Gen Antivirus & EDR (Defender real-time/cloud protection, Controlled Folder Access, Network Protection, exclusions, MDE onboarding, scan and remediation settings from the CIS Defender benchmark)
5 Attack Surface Reduction & Exploit Prevention (ASR rules, DEP/ASLR/CFG/Heap/SEHOP/CET)
6 Application Control & Software Whitelisting (WDAC, AppLocker baseline, installer security)
7 Identity, Privilege Management & Access Control (accounts, LAPS, password/lockout policy, UAC)
8 Network Architecture & Protocol Hardening (firewall profiles and logging, every firewall rule with risky allow rules scored, network exposure map, legacy protocols, RDP, SMB, DNS)
9 Cryptography, Schannel & Secure Channels (TLS/SSL protocol state, cipher suites)
10 PowerShell Subsystem & Execution Security (logging, execution policy, Constrained Language Mode)
11 Advanced Security Auditing & Event Logging (all 27 CIS audit policy subcategories with their exact Success/Failure requirement, process-creation command-line logging)
12 Network Authentication & Credential Exposure Hardening (NTLM, anonymous access, Restrict NTLM, remote UAC for local accounts, RDP Restricted Admin, CredSSP delegation, Kerberos encryption types, LM hashes, stored network credentials, WebClient/WebDAV relay exposure)
13 Privileged Access, User Rights & Session Hardening
14 Legacy Service Exposure & Remote Management Hardening (Print Spooler, WinRM, Remote Registry)
15 Removable Media, App Execution & Modern Threat Surface (SmartScreen, Windows Recall base check)
16 Zero Trust Device Identity, Patch & Time Integrity
17 Advanced Hardware Mitigations & DNS Security (CET, DNS-over-HTTPS)
18 Privilege Escalation — Service, File System & DLL Hijacking (including service permissions that let users reconfigure a service, writable service folders, ServiceDlls and recovery commands, and privileged processes running writable binaries; write access means any non-admin account or a non-admin owner)
19 Credential Theft & LSASS Exposure
20 Privileged Token, Process & COM Hijacking (including privileged scheduled tasks whose program, script argument or folder is writable, and named pipes writable by Everyone)
21 Kernel & Driver Security (vulnerable driver blocklist, test signing, signature validity)
22 Persistence Mechanisms & Unauthorized Auto-Runs (machine-wide autorun targets writable by standard users, AppCertDlls, port monitors, time providers, netsh helpers, password filters, BootExecute, SilentProcessExit, Run keys, Winlogon, WMI subscriptions, autostart inventory, PowerShell profile scripts and their write access)
23 Active Exploitation & Security Subversion Controls (UAC bypass IOCs, event log ACL tampering)
24 AppLocker ACL & Bypass Vulnerability Audit (enforcement, writable paths, LOLBAS, registry/service integrity, Electron apps usable as a Node.js runtime)
25 Modern Identity, Tokens & Cloud Sync (Entra ID PRT/TPM binding, Seamless SSO, WHfB, Intune/MDM integrity, Group Policy vs MDM policy sources)
26 Recoverability, Shadow Copies & Ransomware Preparation (VSS, WinRE, System Restore)
27 Advanced Network Interface, Cellular & Wi-Fi Security (Wi-Fi profiles and the current connection, Mobile Hotspot)
28 Containerization, WSL & Hyper-V Escape Vectors
29 Insecure Temporary Files, Unattended Scripts & Secret Leaks (SSH keys, AWS/Azure CLI, PSReadLine history, passwords saved by PuTTY, WinSCP, FileZilla, mRemoteNG and VNC)
30 Peripheral, Thunderbolt & Media Controls (connected USB and Bluetooth devices, Removable Storage Access, device installation restrictions, Defender Device Control)
31 Artificial Intelligence & Telemetry Hardening (Copilot, Recall, diagnostic data level)
32 Targeted Gap Checks — Defender, Network, Accounts & Awareness
33 Targeted Gap Closure — Credential, Kernel, Network & Privacy
34 System Inventory & Asset Information (process and service inventories, end-of-life runtimes)
35 Browser & Application Hardening (browser extension inventory and risk review for Chrome, Edge, Brave, Vivaldi, Chromium and Firefox; Edge and Chrome security policies; Office macro and Protected View policies)
36 CIS Benchmark Policy Settings (about 290 settings of the CIS Microsoft Windows 11 Enterprise Benchmark v5.0.1 that no other check reads: security options, user rights, account policies, services and Administrative Templates, one finding per area)

Framework Mappings

Every finding carries mappings to:

  • CIS Controls v8 (Safeguard IDs)
  • NIST SP 800-53 Rev.5 control families
  • MITRE ATT&CK technique IDs (v14+)
  • ISO/IEC 27001:2022 Annex A controls
  • NIS2 Directive Article 21(2) risk-management measures
  • NSA/CISA hardening guidance references

These are provided as a practical starting point for compliance documentation and threat-modeling context. They are not a substitute for a formal compliance assessment against the authoritative source documents.


The Remediation Script

Remediate-EndpointPostureCheck.ps1 is generated from the current run's FAIL findings only, grouped by category and sorted by severity. Each block includes the original finding's severity, current/expected value, and framework mappings as comments, followed by:

  • Real, executable commands where automated remediation is safe and unambiguous (e.g. registry hardening values, service state changes, Set-MpPreference calls).
  • Explicit manual-guidance warnings where blind automation would be risky — Active Directory/local group membership changes, User Rights Assignment edits, AppLocker/WDAC policy authoring, and anything flagged as a possible active compromise indicator (which also instructs you to capture forensic evidence before remediating).

The script supports -WhatIf for a dry run and requires -Force to actually apply changes — nothing executes by accident. Always review the generated script before running it in production, and pilot it on a representative device before broad deployment.


Design Principles

A few choices worth knowing about if you're extending this tool:

  • Honesty over false precision. Where Windows doesn't expose a clean, reliable API for something (exact Secure Boot certificate expiration, precise WSL/MDE sensor integration state, CPU mitigation bitmask semantics), the corresponding check says so explicitly and points to the authoritative manual-verification path, rather than asserting a judgement it can't actually back up.
  • Defaults aren't automatically findings. Several checks (SeImpersonatePrivilege assignment, UAC installer detection, Windows Copilot policy) deliberately avoid flagging normal Windows-default or environment-dependent behavior as a misconfiguration — the goal is a low-noise, accurate report, not maximum finding count.
  • No duplicate/contradictory checks. Where two possible checks would evaluate the same underlying setting with different thresholds, only one was kept, to avoid the report disagreeing with itself.
  • Performance-bounded scans. File/log searches (developer secret sweep, third-party agent secret scan) are capped to the most relevant/recent files rather than recursing entire directory trees.

Performance Notes

A full run typically takes 60–90 seconds depending on hardware and installed feature set (WSL, Hyper-V, AppLocker, and Wi-Fi profile checks each add a few seconds). All 180 checks run sequentially in a single process for fault-tolerance and predictable console output ordering.


Responsible Use

This tool includes deep-inspection checks — LSASS/credential exposure, ACL/privilege-escalation vectors, AppLocker bypass surfaces, GPP/SSH/cloud-CLI secret discovery — of the same kind used by tools such as WinPEAS, Seatbelt, and PowerUp. These are defensive, read-only audit checks intended to help you find and fix your own weaknesses before someone else does.

  • Run this only against systems you own or are explicitly authorized to assess.
  • Findings in categories like Credential Exposure & Secret Leaks and Privilege Escalation can surface sensitive material (file paths, partial credential indicators, registry contents). Treat the generated HTML/JSON reports as sensitive documents and store/share them accordingly.
  • The tool performs no exploitation, no credential extraction, and no modification to the target system during an audit run — the only file it writes (besides its own reports) is the optional remediation script, which itself does not execute automatically.

EndpointPostureCheck for macOS

A comprehensive, standalone zsh security posture auditor for macOS 15 (Sequoia) and 26 (Tahoe) endpoints.

Invoke-MacPostureCheck.sh runs 221 read-only security controls across 32 hardening domains — from Secure Enclave/T2 platform integrity and System Integrity Protection through Gatekeeper, TCC privacy controls, FileVault, credential exposure, and ransomware resilience — and renders the results as a single-file, interactive HTML dashboard, in the same visual language as the Windows edition. Every finding is mapped to CIS Controls v8, NIST SP 800-53 Rev.5, MITRE ATT&CK, ISO/IEC 27001:2022, the NIS2 Directive, NSA/CISA guidance, the CIS Apple macOS Benchmark, and the macOS Security Compliance Project (mSCP) — two mappings the Windows edition doesn't carry, since they're macOS-specific authoritative baselines. Findings backed by a real item list (installed apps, listening ports, admin group members, TCC grants, and more) show a popover in the HTML dashboard with the full list, one item per line, instead of cluttering the table with a long inline value. Both editions share the same dashboard (Låda Glass design, sidebar domain filter, KPI filters, theme switch).

It runs standalone with zero external dependencies: no Homebrew packages, no Python, nothing beyond the zsh shell and command-line tools Apple ships on every Mac.

Requirements

Requirement Detail
OS macOS 15 (Sequoia) or macOS 26 (Tahoe). Older versions still run, with one WARN finding noting results may be incomplete.
Shell zsh (the default interactive shell on macOS since Catalina; bash's stock macOS build is too old — 3.2 — for this tool's needs)
Privileges root (the script enforces this at startup and exits immediately if not run via sudo, matching the Windows edition's #Requires -RunAsAdministrator)
Hardware Apple Silicon and Intel both supported. Checks specific to one architecture (Secure Enclave boot-policy checks on Apple Silicon; T2/firmware-password checks on Intel) report INFO "not applicable" on the other rather than a false result.
Dependencies None — no brew install required
Network None required — every check is offline-capable, including patch-compliance checks (which read local softwareupdate --history, never trigger a live -l network scan)

Quick Start

# Run a full audit, write the HTML dashboard to the current directory
sudo ./Invoke-MacPostureCheck.sh

# Full audit + JSON export + generated remediation script, written to a specific folder
sudo ./Invoke-MacPostureCheck.sh --report-path ~/Reports --export-json --generate-remediation --generate-profile

The script must be run with sudo. If it isn't, it exits immediately with a clear message rather than continuing with a partial, misleading result — checks that need a specific console user's preferences (screen lock, Safari settings, iCloud/AirDrop state, etc.) are still read correctly under sudo by dropping to that user for just those reads, not by running the whole audit unprivileged.

Parameters

Parameter Description
--report-path <dir|file> Optional output location for the HTML dashboard. Accepts a directory (default filename is used) or a full file path. Defaults to the current working directory.
--export-json Exports the complete structured finding set, plus posture/summary metadata, to a JSON file alongside the HTML report.
--generate-remediation Generates Remediate-EndpointPostureCheck.sh, containing a remediation block for every FAIL finding from the current run.
--generate-profile Generates EndpointPostureCheck-Hardening.mobileconfig with the compliant value of every setting that a configuration profile can enforce and that came out WARN or FAIL. Written for review and upload to an MDM, never installed.
--baseline <file.json> Compares with an earlier --export-json report: new autostart items, admins, permissions and extensions are a scored WARN; controls that got worse or better are listed.
--exceptions <file.json> Accepted risks, same format and behaviour as the Windows -ExceptionPath (see Accepted risks).
--help Show usage and exit.

Output Files

Same three-file pattern as the Windows edition: EndpointPostureCheck_<hostname>_<timestamp>.html (always generated), EndpointPostureCheck_<hostname>_<timestamp>.json (with --export-json), Remediate-EndpointPostureCheck.sh (with --generate-remediation, supports --dry-run/--force, defaults to preview-only). With --generate-profile there is a fourth: EndpointPostureCheck-Hardening.mobileconfig, an unsigned profile with a Restrictions payload and a Custom Settings payload. Review every setting and test it on a pilot Mac before deploying.

Understanding the Report

Identical PASS/FAIL/WARN/INFO status semantics, Critical/High/Medium/Low/Informational severity levels, weighted compliance scoring (severity weights Critical=10, High=7, Medium=4, Low=2; PASS full credit, WARN half credit, FAIL none; INFO excluded entirely), and LOW/MEDIUM/HIGH/CRITICAL risk badge thresholds as the Windows edition — see that section above for the full breakdown; the scoring engine is a direct, verified port.

Audit Domains

221 checks organized across 32 domains:

# Domain
1 Hardware, Firmware & Platform Integrity (Secure Enclave / T2)
2 SIP & Kernel Security
3 Volume Encryption & Data-at-Rest (FileVault)
4 Gatekeeper, Notarization & Application Control (including disabled download quarantine)
5 Malware Defense (XProtect / MRT / EDR)
6 TCC & Privacy Controls (including powerful grants to terminals/interpreters and grants not pinned to a code signature)
7 Identity, Accounts & Privilege Management (including sudo environment and drop-in permissions, sudo rules that allow shell escapes, weakened authorization-database rights, PAM integrity)
8 Network Firewall & Packet Filtering (application firewall app rules with signature and interpreter checks, pf rules and anchors, network content filters, network exposure map)
9 Remote Access & Sharing Services (including legacy VNC passwords)
10 SSH Server & Client Hardening (including risky sshd options)
11 Cryptography & Certificate Trust
12 Audit, Logging & Telemetry
13 Software Update & Patch Compliance
14 MDM, Profiles & Supervision (CIS macOS Intune payloads: FileVault, passcode, restrictions, Gatekeeper, profile removal)
15 Persistence Mechanisms & Auto-Runs
16 Privilege Escalation — File & Directory Permissions (including ACL-aware writability and running root processes whose binary another account can modify)
17 Dylib Hijacking & Code Injection (including Electron fuses and system-wide library validation)
18 Credential Exposure & Secret Leaks (including root and other users' shell/database histories)
19 Screen Lock & Physical Access
20 Removable Media & Peripheral Control (accessory approval, Thunderbolt, connected USB devices, unencrypted external volumes)
21 Wireless & Network Interface Security (current and known Wi-Fi networks, open networks with auto-join, Bluetooth devices and discoverability, AirDrop)
22 iCloud & Cloud Sync Surface
23 Browser & Application Hardening
24 Ransomware Resilience & Recovery
25 Third-Party Software & Package Inventory (including end-of-life runtimes)
26 System Services & Daemon Exposure (process inventory, launchd services, network connections, third-party services listening on the network)
27 Kernel & Boot Security
28 Virtualization & Container Escape Surface (including exposed container engine sockets and the unauthenticated Docker TCP API)
29 Data Loss Prevention & Sensitive Data
30 System Inventory & Asset Information
31 Apple Intelligence, Siri & Analytics
32 Power Management & Sleep

CIS Apple macOS 26 Tahoe Benchmark v1.0.0 Coverage

Every Automated recommendation in the CIS Apple macOS 26 Tahoe Benchmark v1.0.0 has a corresponding check, tagged with its Tahoe section number (e.g. CIS-macOS 2.6.8) and, where one exists, its mSCP rule ID. Most of the benchmark's Manual/"Audit" recommendations (Find My Mac, Touch ID, Internet Accounts, …) are not scored, since they need human judgement rather than a deterministic CLI signal. The Manual ones that do have such a signal are checked: Lockdown Mode (INFO), all data volumes encrypted (5.3.1/5.3.2), filename extensions shown (6.1.1), the FileVault standby key policy on Intel laptops (2.10.1.1), and a few that score PASS/WARN/FAIL (2.1.1.1, 2.1.1.2, 2.1.1.6, 2.6.2.1).

A few behaviors worth knowing:

  • MDM-aware preference reads. Most CIS settings are read through NSUserDefaults (via osascript -l JavaScript, as root) — the benchmark's own audit method — so values enforced by a configuration profile are seen. A plain defaults read /Library/Preferences/... cannot see profile-enforced keys.
  • Unset profile keys score WARN, not FAIL. About 25 CIS controls (Apple Intelligence, Siri, Safari privacy, AirPlay Receiver, Terminal Secure Keyboard Entry, …) can only pass when an MDM profile enforces them. When the key is simply unset, the finding is WARN ("not enforced by a configuration profile", half credit); an explicitly insecure value is still FAIL. Keys whose CIS audit itself passes when unset (software-update deferral, Time Machine AutoBackup) report PASS.
  • Deliberate deviations from the CIS audit text, each verified on a live Tahoe system:
    • 2.10.1.2 Sleep reads the Battery Power: section of pmset -g custom. The benchmark's pmset -b -g prints the currently in use profile, which is the AC profile whenever the laptop is plugged in.
    • 2.11.1 Screen saver inactivity requires 1 ≤ idleTime ≤ 900. The benchmark's JavaScript audit also passes an unset or 0 ("never") value.
    • 5.2.8 Password history requires ≥ 24, per the recommendation title and mSCP. The benchmark's audit one-liner only tests ≥ 1.
    • 4.3 NFS server: nfsd reporting "enabled" with no /etc/exports and nothing running (the stock state) is a WARN; a running server or an exports file is a FAIL.
  • Time Machine backup encryption is now a real PASS/FAIL check, using each destination's LastKnownEncryptionState (the CIS 2.3.4.2 method), instead of an informational note.

CIS benchmarks beyond the macOS one

  • Chrome and Edge (CIS Google Chrome v3.0.0, Microsoft Edge v5.0.0): the security policies (Safe Browsing/SmartScreen without click-through, download restrictions, no certificate-error override, extension block list, remote debugging off, site isolation) on both editions. A browser with no policy gets one WARN, and on macOS only a value forced by a configuration profile counts.
  • MDM payloads (CIS Apple macOS 26 Tahoe Intune Benchmark): FileVault, passcode, restriction, Gatekeeper, system-extension and profile-removal settings from the installed profiles, on MDM-enrolled Macs only.

Persistence & Autostart Coverage

Domain 15 inspects everything that starts automatically. It covers all local accounts, not just the logged-in user.

  • launchd items (/Library/LaunchDaemons, /Library/LaunchAgents, every user's ~/Library/LaunchAgents). The actual launch target is resolved: when an item runs /bin/bash, python3, osascript, /usr/bin/env … or sh -c, the script or inline command is analyzed, not the Apple-signed interpreter. Each item is checked for:
    • unsigned or ad-hoc-signed targets
    • targets missing from disk (orphaned items, reported as WARN rather than as signature failures)
    • targets or scripts under /tmp or /Users/Shared
    • hidden plists
    • DYLD_INSERT_LIBRARIES set in the plist
    • com.apple.* labels on non-Apple-signed targets (masquerading)
    • LaunchDaemon plists, programs or scripts that a non-root account can modify (root privilege escalation)
  • Login items and background tasks (BTM): enabled login items and SMAppService agents/daemons from sfltool dumpbtm. Paths stored relative to the parent app are resolved, and items the user has switched off are skipped. The dump is loaded once per run.
  • Shell startup files: /etc/z*, /etc/profile, /etc/bashrc, every user's zsh/bash dotfiles (including a relocated ZDOTDIR), and any file they source by literal path. This catches the XCSSET ~/.zshrc_aliases technique and the BlueNoroff ~/.zshenv technique. Both content and permissions are checked.
  • Wider inventory: authorization plugins, DirectoryServices plugins, privileged helper tools, cron and at jobs, rc.local / launchd.conf / StartupItems / periodic / emond, ~/.ssh/rc / sshrc, SSH authorized_keys, and Python .pth / sitecustomize startup hooks.
  • Autostart Inventory (Apps & Scripts): an INFO baseline listing every autostart item across all of the above, classified as program or script, with its signer.

Script content is matched against two pattern tiers. Strong patterns produce FAIL: download-and-execute (curl … | sh), base64-decoded execution, reverse shells, fake password prompts (osascript … with hidden answer), keychain dumping, dscl -authonly password checks, and DYLD injection. Weak patterns produce WARN for review: osascript, nohup / backgrounding, launchctl load, /tmp and /Users/Shared paths, and quarantine stripping. Credential-looking values in matched lines are replaced with [REDACTED], and SSH keys are identified by their comment, never by key material.

What Didn't Port, and What's New

The two platforms don't share a threat surface one-for-one. Dropped entirely (no macOS equivalent): VBS/HVCI/Credential Guard/System Guard, BitLocker specifics (replaced wholesale by the FileVault domain), ASR rules, WDAC/AppLocker's 11-check module (replaced by the 6-check Gatekeeper/notarization domain), LAPS, UAC-specific checks (folded into sudoers/authorization-adjacent identity checks), Schannel registry policy, PowerShell subsystem security, Print Spooler, WPAD/LLMNR/NetBIOS, COM CLSID hijacking, IFEO, AppInit_DLLs, WSL/Hyper-V/Windows Sandbox/WDAG, Windows Recall/Copilot. Added, with no Windows counterpart: TCC & Privacy Controls (Domain 6), iCloud & Cloud Sync Surface (Domain 22), Browser & Application Hardening (Domain 23), Data Loss Prevention & Sensitive Data (Domain 29), and Dylib Hijacking & Code Injection (Domain 17, the closest macOS analogue to DLL hijacking but a genuinely different mechanism).

Framework Mappings

The same six frameworks as the Windows edition (CIS Controls v8, NIST SP 800-53 Rev.5, MITRE ATT&CK, ISO/IEC 27001:2022, NIS2 Article 21(2), NSA/CISA guidance), plus two macOS-specific mappings: CIS Apple macOS Benchmark section references and macOS Security Compliance Project (mSCP) rule IDs — the authoritative macOS hardening baselines, maintained by NIST/NASA/DoD for mSCP, that a macOS-focused auditor is most likely to be asked for directly.

Design Principles

Same three principles as the Windows edition, carried over deliberately: honesty over false precision (several macOS checks — FIPS crypto-module validation, for example — say so explicitly rather than guessing, where macOS exposes no clean, documented CLI signal); no duplicate/contradictory checks (a few candidate checks were cut during design specifically to avoid two checks silently disagreeing about the same underlying setting); performance-bounded scans (shell-history and sensitive-filename sweeps are capped to recent/top-level content, never a full recursive disk walk).


License

Released under the MIT License.

About

A comprehensive, standalone PowerShell security posture auditor for Windows 11 Enterprise endpoints.

Topics

Resources

Security policy

Stars

6 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages