CIS Benchmarks vs DISA STIGs: A SecOps Comparison

Healthcare IT

Choosing a hardening baseline source looks, from a distance, like a small technical decision. In practice it behaves more like a differential diagnosis: the presenting complaint is identical across clinics — “tell us which settings to enforce” — and the right answer depends on the estate in front of you, not on the loudest vendor. This comparison examines the two sources most clinics meet first, CIS Benchmarks and DISA STIGs, and judges them the way The Ops Atlas would: as inputs to a SecOps working contract between security and operations, not as stand-alone purchases. The verdict, stated up front, is that both sources are sound and neither is the decision that matters. The governance wrapped around the chosen source — one owned queue, dated exceptions, staged enforcement — decides whether the baseline ever becomes real.

The contenders, described precisely

Three named sources and one guidance document dominate the field:

  • CIS Benchmarks: a catalogue exceeding a hundred titles, one per technology, from Ubuntu to dental imaging software platforms. Level 1 marks the prudent floor; Level 2 hardens further at a usability cost.
  • DISA STIGs: the United States Defense Information Systems Agency’s configuration guides, the traditional choice in defence-adjacent and government-adjacent environments.
  • The Microsoft Security Compliance Toolkit: Redmond’s own baselines and tooling, aimed squarely at Windows fleets.
  • NIST SP 800-70 Rev. 5, refreshed in May 2026: not a baseline itself, but the guidance governing how security checklists are written, tailored, and maintained.

Underneath all four sits configuration-management doctrine — SP 800-128 carries the lifecycle and SP 800-40 Rev. 4 the patching practice — which every source assumes you will follow. The lifecycle is identical regardless of source: select, tailor (recorded, in XCCDF terms, as a tailoring file), test on a stand-in system, observe quietly in log-only form or within a trial group, enforce in rings, monitor drift, review yearly.

Coverage compared: breadth versus authority

CIS’s advantage is breadth. When a clinic’s estate spans Windows servers, Ubuntu machines, network gear, and cloud tenants, a single publisher with per-technology benchmarks keeps the programme coherent, and the Level 1 / Level 2 split gives a defensible way to stage adoption — floor first, hardened settings for the crown jewels. STIGs’ advantage is authority and depth on the platforms they cover; where a STIG exists, it is exhaustive and battle-tested, and its findings carry weight with government-adjacent assessors. The honest weakness of each mirrors the other: CIS breadth can outrun its automation support on niche platforms, while STIG coverage simply stops at the agency’s perimeter of interest. A clinic running mixed estates should map every asset class against both catalogues before choosing, the way a clinician rules a diagnosis in or out before treating. Record the mapping, too — the gaps you document today become next year’s procurement questions instead of next year’s surprises.

Grading: who actually checks SecOps conformance

Whichever source wins, someone must grade the estate against it. The recognised graders: CIS-CAT Pro, published by CIS itself; OpenSCAP, usually paired with the ComplianceAsCode content library; Ubuntu’s Security Guide for its own platform; and the SCA module inside Wazuh. Results here demand the same scepticism a careful clinician brings to a single lab value. A conformance scan is a snapshot of settings on one afternoon; it demonstrates nothing about how enforcement behaves under a Monday-morning load, which is precisely why the lifecycle insists on observation before mandate. Microsoft’s Netlogon campaign for CVE-2020-1472 remains the canonical demonstration: phase one of enforcement on 11 August 2020, denials and borderline connections surfaced through events 5827, 5828, and 5829, permitted ones through 5830 and 5831, an allow-list that had to reach empty ahead of the irreversible step on 9 February 2021. Green scan, staged enforcement, empty allow-list — in that order.

The shared SecOps weakness: version drift

Both sources drift, and drift is where programmes quietly rot. The Atlas’s cautionary pairing: CIS Benchmark v2.0.0 for Ubuntu now exists alongside automation content that still tracks the older v1.0.0/2024 editions, so a scanner may grade an estate against rules nobody currently recommends. The lesson generalises to any source, STIGs included: pin the exact edition in the register, reconcile scanner content against it, and re-check at every yearly review. A version number in the minutes is worth more than a colour on any dashboard ever built.

Where both sources depend on the SecOps contract

Now the part neither source can do. A baseline enforces nothing by itself; it needs the machinery around it:

  1. A single remediation queue, so every scan finding lands with one owner and one date — platforms such as ServiceNow’s Yokohama release automate the assignment through the CMDB.
  2. Resourced timeframes per owner group, because NIST treats the estate-wide patch percentage as not actionable; the September worked example (82.5% overall, 95% versus 75% across cells) shows the average concealing the failing cell.
  3. Governed exceptions with expiry, counted weekly — six of thirty-five past review in the same example.
  4. Telemetry duties owed the watchers under CIS Control 8, from logging basics at 8.1-8.3 up to IG2’s ninety-day retention and its weekly review obligation.

Swap either source out and those four obligations survive unchanged. That is the tell: the baseline source is a variable; the contract is the constant. It is also the cheaper half of the decision, since governance costs discipline rather than licence fees. The Atlas’s full framing of the contract, including which duties stay out of scope, lives at its SecOps entry.

The verdict for clinic IT

Score the comparison honestly and it ends level, which is the clinically correct result. Choose CIS Benchmarks for breadth and the Level 1-2 staging path; choose STIGs where authority or a covered platform demands it; add Microsoft’s toolkit where Windows dominates the estate. Then spend your remaining energy on the governance layer, because that is where programmes succeed or quietly die. Readers who want the underlying habit — evidence over instinct, structured comparison before treatment — will recognise it from recognising bronchitis by its early signs and from the general discipline of knowing when respiratory symptoms need medical help. Diagnosis first, baseline second, enforcement staged third — the order holds whether the patient is a person or an entire estate, and structured online consultations prove the same point in another domain: good inputs, clearly owned, beat heroic improvisation every single time.