Skip to content
Documentation

Discovery Scans

Discovery Scans turn files observed on real endpoints into policy candidates.

Discovery Scans turn files observed on real endpoints into policy candidates. Instead of inventing an allowlist from a vendor list, UEC asks a device what it actually runs and then proposes rules at the trust level you choose.

Discovery scans in light mode Discovery scans in dark mode


Situation Good scan source
New department baseline Recent blocks and audited executions
Portable software not in registry Folders
Build a policy from live policy telemetry A policy’s observed events
Software Catalog entry needs trust Catalog-linked scan for that device
Blocked execution requests JIT JIT-linked scan/request
One-off evidence from a saved log EVTX upload / event source where supported
Known folder of tools Explicit folder list

  1. Open Discovery Scans.
  2. Select New Scan.
  3. Pick an eligible device. Offline devices can be selected; ineligible devices explain why.
  4. Name the scan and state whether results are for Trust or Block.
  5. Select a source:
    • recent blocks;
    • folders;
    • events since a date;
    • a linked software-catalog or JIT entry where applicable.
  6. Add up to 50 folders when using folder discovery.
  7. Choose a rule level and observe its fallback chain.
  8. Set an identity limit if you need a bounded scan.
  9. Submit.

New discovery scan in light mode New discovery scan in dark mode

The form only offers rule levels the platform and source can build. A macOS scan cannot build a block policy; options that do not apply are disabled rather than submitted and rejected later.


Rule generation prefers stable cryptographic identity:

Publisher / signer → path / PFN → Authenticode hash

A fallback is not a silent “anything goes”. It records what identity was available. Hash-only trust should be reserved for a binary you deliberately want to pin.


The scan detail page shows:

Field Meaning
Queued Request accepted by the server
Sent Agent received the command
Receiving Results are streaming back
Completed / failed / cancelled Final state
Events read Input events considered
Files identified Files resolved to identities
Entries Candidate rules received
Last change / last checked Server progress and page refresh

The page polls while a scan runs and displays partial results when they are available. Decisions open after completion.

A cancelled scan can still finish on the device because the agent has no stop command; late results are discarded by the server. Do not assume cancellation reclaims all device I/O.


  • Search entries by path, file, signer or identity.
  • Filter by Kind and Decision chips with counts.
  • Sort by order found, entry or rules.
  • Open a row to see all rules generated from one entry.
  • Select rows or select all matching entries.
  • Include a candidate in policy generation.
  • Exclude noise before building.

The result state is kept in the URL beside policy, intent and build-job identifiers, so a reload does not lose your selection.


  1. On a completed scan, select Build a Policy.
  2. Name the policy.
  3. Choose Trust or Block where the platform supports it.
  4. For a deny policy, choose audit or enforce.
  5. Select which entry kinds to keep.
  6. Optionally restrict a Windows policy to kernel mode.
  7. Confirm the dialog. It names the policy, scan and number of entries.
  8. Follow the queued build job; open View policy when it completes.

A signer block is especially broad: it can block everything signed by that publisher. The confirmation tells you this before you commit.


  • Scan a representative device, not necessarily the loudest one.
  • Prefer signer/publisher rules for maintained software.
  • Prefer hashes for fixed tooling or suspicious one-offs.
  • Exclude updater caches, temp folders and known-noisy paths before building.
  • Do not mass-authorize a publisher solely because prevalence is high; verify the publisher is expected in your environment.
  • Keep scan names linked to the business unit or incident.

  • One running scan per device is not enforced; the agent may run concurrent scans.
  • A cancelled running scan still finishes locally.
  • Legacy result rows without per-row entries can be read and built from, but not individually included/excluded.
  • A scan is evidence from one device and moment; combine it with Software Catalog and File Identities for fleet context.