What does a LOLBAS entry's Detection block give a detection engineer, and where does it stop being reliable?
answer
- six subfields, mostly links
- free text next to rule links
- the project disclaims coverage
- legitimate admin use looks the same
- some entries have none
basics
~20 sThe Detection block lists IOC items and links to Sigma, Elastic and Splunk rules, BlockRule policies and Analysis write-ups. It is a set of starting points: the project does not guarantee that any of it detects the behaviour.
solid answer
~40 sA `Detection` block can hold free-text `IOC` items and links to community `Sigma`, `Elastic` and `Splunk` rules, a `BlockRule` link (usually Microsoft's recommended block rules) and `Analysis` write-ups. As read on 2026-10-04, 232 of 248 entry files have at least one item, and Sigma is the most common. The project's notice says it does not guarantee that a referenced rule or IOC will detect the behaviour, and does not claim complete coverage. So I treat each item as a candidate: check it fits my log source, baseline it against legitimate administration to see the false-positive rate, test it in an authorised exercise, and give it an owner. Entries with no block need a detection built from process-creation data.
go deeper
Recall the Detection subfields: IOC, Sigma, Elastic, Splunk, BlockRule and Analysis, and that they are references, not finished protections.
Explain that most items are links to community rules, that BlockRule usually points at Microsoft's block rules, and that some entries have none.
Demonstrate the validation chain: log-source fit, false-positive baseline, authorised test, owner. Quote the project's own disclaimer when asked whether the references are enough.
Describe how a detection programme tracks catalogue-derived coverage honestly: which entries are covered, by which tested rule, and which are known gaps reported to the owner.
## What is in the block The template and the project's notice define the subfields: - **`IOC`:** free-text indicators, such as a process relationship, a file or registry artefact, or a statement that the file should not run on a normal workstation. Quality varies from entry to entry. - **`Sigma`, `Elastic`, `Splunk`:** links to detection rules held in those projects' repositories. - **`BlockRule`:** a link to a blocking policy. Of the 24 entries that carry one, 22 point to Microsoft's recommended block rules, one to Microsoft's page on applications that can bypass application control, and one to a Sigma rule. - **`Analysis`:** links to write-ups. Only four entries use it. ## What the data shows Counting the 248 entry files as read on 2026-10-04: | Item | Files containing it | |---|---| | any Detection item | 232 | | `Sigma` | 204 | | `IOC` | 155 | | `Elastic` | 51 | | `Splunk` | 33 | | `BlockRule` | 24 | | `Analysis` | 4 | Another 15 files have no `Detection` key at all, and one has the key but no items. Contributors add references when they know of them, so the spread follows who wrote about what, not what is detectable. ## Where it stops being reliable 1. **The project disclaims it.** The notice says LOLBAS does not claim its detection references give complete defensive coverage, and does not guarantee that a particular rule or IOC will detect the behaviour. A link is an acknowledgement that a resource exists. 2. **Links are not pinned or reviewed by you.** A rule can change or disappear upstream; one `BlockRule` link even carries a commit hash. Copy what you use into your own repository and record the source and date. 3. **A rule covers what it was written for.** A rule keyed to one command shape or one file name covers that shape. Treat it as a hypothesis about the function and test whether it fires on the variants your own review finds. How such a rule is written belongs to the detection-engineering material. 4. **Legitimate use looks the same.** The files are signed and used by administrators, installers and agents. A rule that fires on the file alone will be noisy, so baseline it against your fleet before it pages anyone. 5. **Fit with your telemetry.** A rule only helps if your logs carry the fields it relies on. Checking that, and mapping fields where they differ, is a separate piece of work. ## A workable process 1. Select the entries that exist on your hosts and matter for the host role. 2. Pull each reference into your own repository, with the entry's date and source. 3. Check it parses and that the fields exist in your logs. 4. Run it over historical data to measure the false-positive rate. 5. Test it in a lab or authorised exercise, so you know it fires. 6. Assign an owner, a severity in your own scale, and a review date. For entries without detection items, start from process-creation data and the entry's category, and say so in the coverage report instead of showing the entry as covered. ## Choosing between the subfields - **`BlockRule`** answers "should this run here at all?" and belongs with the application-control owner. It is policy guidance, not an alert. - **`Sigma`, `Elastic`, `Splunk`** answer "did it run in a suspicious way?" and belong with the detection engineers. Prefer the format your platform consumes natively and treat the others as reading material. - **`IOC`** is useful for hunting and for triage notes, but its free text rarely drops straight into a rule. - **`Analysis`** links are context for the analyst. Nobody maintains them for you. ## What to report Report coverage honestly: the entries with a tested rule, the entries with a candidate rule awaiting a baseline, and the entries with no reference. Give the date and source of each reference, and say which were validated in an exercise. A coverage figure that counts every linked rule as deployed overstates protection, and the project's own notice says as much.
- An entry has a Sigma link. Can you deploy that rule to production as it stands?Not without work. Check that your log source has the fields it uses, run it against historical data to see how often it fires on legitimate administration, and test that it fires in a controlled exercise. The project does not guarantee the rule detects the behaviour. Conversion and field mapping are their own task.
- What does an entry with no Detection items mean for a defender?Only that no reference was contributed. It does not mean the use is undetectable or unimportant. Build detection from process-creation telemetry and the entry's category, record in your coverage report that the entry has no community reference, and revisit when the catalogue is updated.
- Every Detection item is tested and maintained by the project
- A Sigma link in an entry means the behaviour is fully covered
- BlockRule links are policies that ship enabled with Windows
- An entry with no Detection block is undetectable
- Deploy all linked rules and the catalogue is covered
- IOC items are verified signatures that need no tuning