Why did ATT&CK give ICS techniques their own T0### range, and why does a T1### id no longer rule ICS out?
answer
- three domains, not one tree
- columns differ, not just contents
- no cross-matrix parent or child
- Mobile is separate yet still T1###
- most ICS ids are T0###, not all
basics
~20 sBecause ICS is a separate matrix, not a branch of Enterprise, with its own tactic columns and technique list, so its techniques were numbered in their own space. A T0### id still means ICS, but since v19 some ICS techniques carry T1### ids too, so the separation lives in the columns, not the digits.
solid answer
~50 sATT&CK publishes several domain matrices — Enterprise, Mobile and ICS — and each is its own coordinate system: its own ordered tactic columns and its own techniques. ICS carries goals that Enterprise has no column for, such as Impair Process Control and Inhibit Response Function, because Enterprise has no notion of a physical process to mis-drive. The `T0###` range made the non-nesting visible: no ICS technique is a child of an Enterprise technique, and you cannot roll ICS numbers up into Enterprise or add the two technique counts together. Most ICS techniques still carry live `T0###` ids, but the digits were never the rule. Mobile is a separate matrix numbered in the same `T1###` space as Enterprise, and since v19 ICS has `T1###` techniques of its own, such as `T1692` Unauthorized Message. A `T0###` id still means ICS, but a `T1###` id could sit in any of the three matrices, so you look the object up.
go deeper
Be ready to say plainly that Enterprise, Mobile and ICS are separate matrices with their own tactic columns, and that ICS techniques got their own T0### range because they are not children of Enterprise techniques, though since v19 some ICS ids are T1### too.
Explain why the columns differ, not just the numbers: ICS carries goals such as Impair Process Control that Enterprise has no cell for, and name an action that Enterprise genuinely cannot express.
Show what the separation forbids in practice — no nesting, no summed counts across matrices, no dedupe on shared technique names — and note that the digits only work one way: T0### still means ICS, but an ICS technique need not be T0###, and a T1### id could sit in Enterprise, Mobile or ICS.
Own the consequence for reporting: any figure quoted against 'ATT&CK' without naming a matrix is not comparable across estates or vendors, and mixed IT/OT estates need two named denominators rather than one blended one.
## ATT&CK is several catalogues, not one tree People say "ATT&CK" as though it were a single list of numbered attacker behaviours. It is not. MITRE publishes several **domain matrices**, and each one is an independent coordinate system made of three things: an ordered set of **tactic columns** (what the adversary is trying to achieve at that moment), a set of **techniques** placed under those columns (how they achieve it), and the identifiers that name them. The domains are Enterprise, Mobile and ICS (industrial control systems). A "domain" here is not a filter or a view. It is a redraw. Two domain matrices can disagree about what the columns even are, because the adversary's available goals differ once the assets differ. ## What the ICS matrix has that Enterprise does not The clearest evidence is the columns. The ICS matrix carries tactics such as **Impair Process Control** and **Inhibit Response Function** — goals with no Enterprise counterpart, because in Enterprise there is no physical process to drive to the wrong value and no safety function to stop from firing. Enterprise's Impact column is about availability and integrity of data and services; it does not stretch to "the pump now runs at a setpoint the operator did not choose". The same asymmetry shows in the techniques. Consider an ICS technique like **Unauthorized Message: Command Message** (`T1692.001`) — sending a well-formed, protocol-legal command that the device simply obeys, because the control protocol's specification contains no authentication at all. Try to place that in the Enterprise matrix. It is not Execution: no attacker code runs anywhere. It is not Lateral Movement: nothing was logged into. It is not Exploitation: nothing was exploited, the device did exactly what its designers said it would. Enterprise has no cell for it, which is exactly why ICS is a matrix and not a platform tag. ## What the separation asserts The separate matrix is a structural claim, one the original `T0###` range made visible, and it forecloses three specific errors: 1. **Nesting.** No ICS technique is a sub-technique of an Enterprise technique, and no Enterprise technique is the "IT version" of an ICS one. There is no parent-child edge across matrices in either direction. 2. **Summation.** "We address 140 techniques" is meaningless if some are Enterprise techniques and some are ICS ones. The denominators are different catalogues, so the numerator is not a quantity. 3. **Comparison.** Two estates measured against different matrices have not been measured against the same thing, even where the technique names look similar. That last point matters because **names recur across matrices**. Several familiar technique names appear in both the Enterprise and ICS catalogues, but each occurrence is a distinct object with its own identifier, its own description and its own asset scope. Deduplicating by name silently merges two different claims. ## The prefix is evidence, not a rule Here is where a confident answer goes wrong. Having learned that ICS is `T0###`, people generalise: every matrix must have its own prefix, therefore the digits identify the domain. They do not, for two reasons. **Mobile is the third domain matrix and its techniques are numbered in the same `T1###` space as Enterprise.** And **since ATT&CK v19, ICS has `T1###` ids too**: that release merged nine `T0###` techniques into five new ICS parents, `T1691` to `T1695`, so v19.2 counts 81 `T0###` and 16 `T1###` ids in the ICS matrix. Identifiers are unique across all of ATT&CK, so there is no collision — but there is also no readable signal. Given a bare `T1###` number you cannot tell from the digits whether it is an Enterprise, Mobile or ICS technique; you have to look it up. So the honest statement is: the `T0###` range was good evidence that ICS is separate, it still covers most ICS techniques, and a `T0###` id still points only at ICS — but separateness is a property of the matrix, not of the numbering scheme. Mobile proves it by being separate without a range of its own, and the new ICS `T1###` parents prove it from the other side: they sit under ICS columns, and their sub-techniques hang only from ICS parents. ## Why interviewers ask this It is a cheap test of whether you have opened the catalogue or only cited it. The wrong answers are all reasonable-sounding: "T0 is the legacy numbering", "ICS is the OT sub-tree of Enterprise", "each matrix has its own prefix". Each one implies a wrong operating model — that ICS behaviours roll up into an Enterprise picture, that an OT-adjacent estate can be described in one coordinate system, that you can add the numbers. The correction is a single fact you either know or do not: they are different matrices with different tactic columns, and nothing nests. ## Where it bites Any estate that is neither purely office nor purely plant — building automation, physical security devices, lab and facility equipment sitting on the same network as laptops — forces the question of which matrix describes it. The answer is normally "both, per asset group, joined by hand", and knowing that the two matrices do not nest is the first step to saying so without pretending the join comes for free.
- If ICS techniques were numbered T0###, what prefix marks Mobile?None. Mobile is a separate domain matrix, but its techniques are numbered in the same T1### space as Enterprise. Identifiers are unique across all of ATT&CK, so nothing collides, but a T1### number carries no domain signal, and since v19 not even ICS is cleanly ranged, because newer ICS parents such as T1692 Unauthorized Message are T1### too. You identify a technique's matrix by looking the object up rather than by reading the number.
- A technique name appears in both the Enterprise and ICS catalogues. Same definition?No. Each is a distinct object with its own identifier, description and asset scope, written for a different set of devices. Treating the shared name as one entry merges two different claims and quietly halves your technique count. If you are joining Enterprise and ICS material, join on identifiers and keep both, never on names.
- Someone reports coping with 140 ATT&CK techniques across a mixed IT and OT estate. What is wrong with the number?It sums across two catalogues. Enterprise and ICS have different tactic columns and different technique populations, so a count that mixes Enterprise and ICS techniques has no denominator behind it and cannot be compared with anyone else's figure. Report two numbers against two named matrices, or report none.
Two maps of the same country drawn on different grids. A grid reference from one is not a coarser or finer version of the other's — it is unreadable there.
- Calls ICS a sub-tree or OT branch of Enterprise ATT&CK
- Reads T0### as an older or deprecated numbering scheme
- Reads a T1### technique's matrix off the digits of its identifier
- Adds Enterprise and ICS technique counts into one total
- Merges same-named Enterprise and ICS techniques as duplicates
context
resolving link…