Code has a weakness taxonomy: CWE names flaw classes, and TLCTC maps them onto ten causes. It is tempting to treat CCE as the same thing for configuration. It is not. A CCE names a setting; in 61% of the legacy entries it does not even say which value is secure. The weakness only appears as a deviation from a baseline value that CIS, a STIG or Apple's mSCP supplies. The list carries no reference to any threat, issues new identifiers for the same rule on every platform release, and has not covered a current Windows version since 2012. Read through TLCTC, a CCE is a control-object identifier: useful as a join key in tooling, never as a classification. What the configuration side lacks is not identifiers but the generic level between 17,604 leaf settings and the ten clusters.
What CCE Is
The Common Configuration Enumeration began at MITRE and is maintained by NIST as part of SCAP, the Security Content Automation Protocol. NIST's definition: CCE provides “unique identifiers to security-related system configuration issues”, so that the same setting can be correlated across tools, checklists and reports. Each entry carries five attributes: the identifier (for example CCE-2715-1), a description, the conceptual parameters, the technical mechanisms that implement it (a registry path, a file, a command), and references into baseline documents.
CCE is not a relic. NIST SP 800-70 Rev. 5, the National Checklist Program guidelines finalised in May 2026, lists among its key highlights “enhanced mapping concepts between checklist settings, NIST Cybersecurity Framework (CSF) 2.0 outcomes, SP 800-53 controls, and Common Configuration Enumeration (CCE) identifiers”. So it is worth asking precisely what a CCE identifies — and what it does not.
What the List Contains
The current release is 5.20251209 (9 December 2025), a combined spreadsheet with one sheet per platform. We read all of it.
| Finding | Value |
|---|---|
| Entries / platform sheets | 17,604 / 54 |
| Sheets that track current platforms | Apple through macOS 26 Tahoe, iOS 26 and visionOS 26 (fed by Apple's mSCP project), RHEL 8, SUSE 12 and 15 |
| Newest Windows coverage | Windows 7 and Server 2008 R2, last modified 2012 |
| Not covered at all | Windows 10 and 11, Windows Server 2012 to 2025, RHEL 9 and 10, Chrome, Edge, cloud services, network devices; the only container platform is OpenShift 3 |
| Legacy entries with value-neutral wording | 7,411 of 12,171 (61%) — “… should be configured correctly” |
| References to CWE, CAPEC, ATT&CK or CVE | none — where entries map to anything, it is SP 800-53, a control catalogue |
A typical legacy entry, CCE-12007-1: “The ‘6to4 Relay Name’ machine setting should be configured correctly.” Parameters: enabled/disabled. The entry tells you which knob; it leaves the secure position to someone else. The modern entries work the other way round — CCE-95354-7 reads “Disable Airplay Receiver”, with the value built in. One namespace, two meanings: a setting, and a rule about a setting.
Test 1: Is a CCE a Cause?
TLCTC's Axiom III says threats are causes, not outcomes; every cluster is defined by the generic vulnerability an attacker exploits. A CWE fits that grammar because a CWE already names something negative: Weak Password Requirements, Use of Default Password, Initialization of a Resource with an Insecure Default. That is why the CWE mapping can place 791 of them in a cluster.
A CCE names a knob. It is neither a cause nor an outcome. The weakness only exists as a deviation: the setting holds a value that a baseline marks as insecure. So the configuration-side counterpart of a CWE is not a CCE but a triple:
CCE (which setting) + baseline value (CIS / STIG / mSCP) + observed deviation
= a weakness that leaves some cluster's generic vulnerability exposedA scanner finding that says “CCE-80654-7 failed” means nothing until you know which baseline it was checked against. The meaning lives in the baseline, not in the identifier.
Test 2: Can a CCE Be Classified?
Axiom VI — one step, one generic vulnerability, one cluster — applies to attack steps. A setting is not a step. It is a control object: it changes whether, or how cheaply, a step can succeed. So a setting is not classified; it is placed, like a NIST CSF outcome in the control matrix, and one setting may serve several rows or none. Following the insecure value to the step it enables shows how different the placements are:
| Rule (CCE) | If the insecure value stays | TLCTC reading |
|---|---|---|
Disable Remote Apple Events (CCE-95392-7, macOS 26) | Anyone with valid credentials can script the Mac remotely through a designed interface | #4 → #1: the setting decides whether the designed function is reachable after credential use |
PAM password requirements (CCE-80654-7, RHEL 8) | Guessable passwords (CWE-521) | #4: guessing is repeated credential application (ATT&CK T1110 maps to #4) |
Failed-attempt interval (CCE-80669-5, RHEL 8) | Unlimited attempts (CWE-307) | #4 |
USB devices must be authorized (CCE-95416-4, macOS 26) | Any plugged-in device or medium is accepted | #8, and #8 → #7 when the medium carries code (T1091) |
Randomized address space layout (CCE-80916-0, RHEL 8) | Memory-corruption exploits get cheaper | No new cause: #2/#3 remain the cause; the setting only raises their cost |
Audit user/group changes (CCE-80761-0, RHEL 8) | Changes go unrecorded | No cluster: detection capability, an umbrella DETECT control for every row |
Disable Personalized Advertising (CCE-95390-1, macOS 26) | Usage data flows to the vendor by design | Outside the threat axis: a privacy decision about a designed data flow, not an attacker's cause |
And when an insecure value breaks something with no attacker involved, it is not a threat at all. The core paper names misconfiguration as a standing example of System Failure: a loss of function that serves no attacker's objective, operational risk, no cluster.
Test 3: Is a CCE a Stable Identifier?
Identifiers earn their keep by staying put. CCE's do not. Between macOS 15 Sequoia and macOS 26 Tahoe, all 316 rules the two releases share received new CCE identifiers — while the mSCP rule identifiers stayed the same for 315 of them. Disable AirPlay Receiver is CCE-91932-4 on Ventura, CCE-94348-0 on Sequoia and CCE-95354-7 on Tahoe; its rule id is system_settings_airplay_receiver_disable on all three.
The same holds across platforms. Recording changes to /etc/passwd is CCE-80761-0 on RHEL 8 and CCE-85577-5 on SUSE 15. A minimum-password-length rule appears 46 times across 30 sheets, with nothing that ties the copies together. CWE has a hierarchy — pillars, classes, bases, variants — which is what lets a thousand weaknesses be reasoned about at ten causes. CCE is flat: 17,604 leaves, re-issued per version, with no level above them.
Where the Maintained Rules Land
We placed the 1,259 rules of the three maintained sheets with a keyword pass — rough by design, about four in five plausible on a spot check — to see the shape, not to publish a mapping:
| Placement | macOS 26 Tahoe | RHEL 8 | SUSE 15 STIG |
|---|---|---|---|
| Audit, logging, time (no cluster: detection) | 19% | 36% | 45% |
#1 function and permission scope | 32% | 27% | 18% |
#4 authentication | 16% | 9% | 16% |
#5 channel and crypto | 5% | 6% | 4% |
#8 unattended device, removable media | 4% | 3% | 4% |
#7 execution | 3% | 3% | 2% |
#10 update sources, signatures | 3% | 1% | <1% |
#2/#3 exploit mitigation | — | 1% | <1% |
| Unplaced by the keyword pass | 17% | 12% | 9% |
Three things stand out. First, the canon holds: configuration's home is #1, whose generic vulnerability the v2.6 dictionary defines as “The inherent trust, scope, and complexity designed into software functionality and configuration.” #4 comes second. Second, a large share of what is called hardening is detection configuration — it serves no particular cause and belongs in the umbrella DETECT column, not in any cluster row. Third, #2 and #3 barely appear. Configuration hardly touches implementation flaws; those are closed by patching, which CCE leaves to CVE and OVAL. That is no contradiction of our CIS Controls v8.1 mapping, where #2 carries the most preventive Safeguards of any cluster (19): the CIS Controls include vulnerability and patch management; configuration rules do not. #6 and #9 do not appear at all — correctly: capacity is engineered, not toggled, and people are not configured.
What the Configuration Side Is Missing
The problem is not too few identifiers; it is the missing middle. On the code side the chain is complete: a finding carries a CWE, the CWE sits in a hierarchy, the hierarchy maps to a cluster. On the configuration side a finding carries a CCE — or, more often today, a baseline's own rule id or a CIS recommendation id — and the chain stops there.
Part of the missing middle already exists, inside CWE. CWE names configuration weaknesses, and the TLCTC CWE mapping already places them:
| CWE | Name | Cluster |
|---|---|---|
| CWE-1188 | Initialization of a Resource with an Insecure Default | #1 |
| CWE-276 | Incorrect Default Permissions | #1 |
| CWE-521 | Weak Password Requirements | #4 |
| CWE-307 | Improper Restriction of Excessive Authentication Attempts | #4 |
| CWE-1393 | Use of Default Password | #4 |
| CWE-319 | Cleartext Transmission of Sensitive Information | #5 |
That suggests a bridge worth testing: configuration rule → the CWE its insecure value instantiates → cluster, through the mapping that already exists. Rules that instantiate no weakness — audit, time, privacy — fall out of the threat axis by the same test, which is exactly where they belong. This is a proposal, not canon; whether it holds across a full baseline is an empirical question.
What This Means in Practice
- Use CCE as a join key, never as the classification. It correlates a setting across a checklist, a scanner and a report. It does not tell you what the setting defends against.
- Prefer the baseline's rule id for stable references. The mSCP rule id, the content-project rule id or the CIS recommendation id survive releases; the CCE does not.
- Report deviations by cluster, not by count. “214 failed CCEs” says nothing about risk; “open
#4and#1exposure on internet-facing hosts” does. Separate the detection-configuration findings — they degrade every row's DETECT cell, not one cause. - Keep patch state out of hardening metrics. Missing patches are
#2/#3exposure, measured against CVEs; configuration rarely moves those rows. - For Windows, the CCE list is not your source. Use the vendor's security baselines or CIS Benchmarks directly.
For a few hours on 1 October 2026 the Framework Position diagram on our landing page listed “CIS • CCE” as development vocabulary, next to CWE and OWASP. This analysis is why CCE came off again. CIS stays: hardening baselines are a language engineering teams really speak. A setting identifier is not.
Further Reading
- MITRE CWE → TLCTC mapping — 1,012 CWE entries, 791 of them placed in the ten clusters, including the configuration CWEs above
- The ten cluster definitions — generic vulnerabilities, verbatim from the v2.6 dictionary
- NIST CSF 2.0 and TLCTC — outcomes placed in the control matrix; the same placement logic applies to settings
- 153 Safeguards, 74 objectives — every CIS Controls v8.1 Safeguard placed in the TLCTC matrix
- What CIS Cannot See — the same Safeguards by cause, and the skew an asset axis cannot show
- SSDLC for developers — the Programmer/Coder split, and where configuration still lacks an owner
- They Patched the Path. The Cluster Came Back in 24 Hours. — why fixing one instance leaves the generic vulnerability
- The TLCTC core paper — Axiom III, Axiom VI, System Risk Event and System Failure
Sources. NIST, Common Configuration Enumeration (CCE) — definition and entry attributes. NIST, National Checklist Program: CCE — the CCE list, release 5.20251209 (combined spreadsheet, 9 December 2025), from which every count, identifier and quotation of an entry in this post is taken. NIST, SP 800-70 Rev. 5 (initial public draft December 2025, final May 2026). The placement percentages come from a keyword pass over rule titles and are indicative only. CWE and ATT&CK placements are quoted from the TLCTC mappings in the framework repository; the #1 generic vulnerability is quoted from the v2.6 dictionary.