Patching
When a hunt finds a host running a vulnerable version of a package, Scout can say which package to update and to what version, and propose a patch. Where no patch Action fits, it drafts a script for a person to review. Patching never runs on its own: an organisation admin approves every patch and publishes every script Scout drafts.
On this page you see how Scout suggests patches, how the patch Actions run, and what happens to a script Scout drafts.
How Scout suggests a patch
Scout suggests patches after a confirming re-hunt, or as soon as a hunt finds a vulnerable package. For each affected host it puts three things together:
- What is installed — the packages and programs the endpoint agent reports: Debian and RPM packages, Windows programs, Homebrew and Python packages.
- What is vulnerable — the CVE, OSV and CISA KEV (Known Exploited Vulnerabilities) data Huntbase holds for those packages.
- The fixed version — the first version that fixes the vulnerability, from the OSV record's affected and fixed ranges.
Suggestions appear as Patch suggestions on the hunt's Brief and in Scout's answers. Each host lists its packages, each with the installed version, the version that fixes it, its CVEs, a KEV badge when one is On CISA's Known Exploited Vulnerabilities list, and what Scout suggests:
| Suggestion | When |
|---|---|
| Propose the Patch Action | A fixed version is known and the package manager can install it: Scout proposes Initiate patch. |
| Propose installing security updates | The operating system's own security updates cover it: Scout proposes Install security updates. |
| Scout drafts a patch script for a person to review | Neither Action fits, on a host the endpoint agent runs on. See below. |
| Patch through the vendor | An appliance or other host without the endpoint agent, such as a firewall or a file-transfer server. Patch it through the vendor's own process; Huntbase never automates this. |
Suggestions are written into the hunt's evidence as checks, not findings, so the hunt's verdict and Scout's answers can cite them.
Scout doesn't guess. When a fix doesn't cover every CVE, the row says Some CVEs have no published fix. CVEs Scout couldn't confirm are listed under Not verified, with the reason. What it couldn't see at all, such as a host with no installed-package list, is listed under Missing data.
Patch Actions
Scout proposes one of two starter Actions, whichever fits:
| Action | What it does |
|---|---|
| Initiate patch | Upgrades one package to at least a minimum version on each endpoint. Reboot is off by default. |
| Install security updates | Installs only security updates, through the operating system's own update mechanism. |
Both are high risk and work the same way:
- An organisation admin approves every run. Patching is never covered by a response policy, however the policy is set up.
- Canary first. The run starts on a small share of the targets and waits for a person to click Continue before the rest. See Rollouts.
- Checked afterwards. Each endpoint is read back once the patch has run: Initiate patch checks the package is at or above the target version, and Install security updates checks no security updates are still pending.
- No rollback. Neither Action has an undo. To go back, pin or reinstall the package by hand.
When no Action fits, Scout drafts a script for a person to review
Sometimes neither Action fits, for example a package installed outside the operating system's package manager. Scout then writes a script that would do the job, and leaves it for a person.
- It's saved as a draft, unpublished, in Content library › Scripts, marked Script drafted by Scout: Scout wrote this draft. A person reviews, tests and publishes it. Its page shows What it is for, the Code, How to check it worked, Rolling it back, and Open the hunt it came from.
- It isn't checked or tested. The draft carries a banner, Not checked or tested: review before use. Huntbase runs no automatic checks on it.
- Scout never runs it. Endpoints refuse unpublished scripts, and Scout can't test or publish one.
- A person is asked. Your organisation admins get an item in Needs you, Scout drafted a script — review before use, and Scout activity records the draft.
Review the draft, and test it the way you would any script. An organisation admin publishes it: Publish vn (the draft's version), then Sign and publish, through the usual review and publish path. Other members see An org admin reviews and publishes scripts Scout drafts. Once published it is an ordinary script: an Action can use it, and any run Scout proposes with it is approved like any other.
What never happens on its own
- Installing a patch or security updates. No response policy can allow it, and Scout only ever proposes.
- Running, testing or publishing a script Scout drafted.
- Patching through a vendor.
Next steps
- Response policies — what Scout may do on its own, and what it never does
- Actions — the starter patch Actions and their approval
- Scripts — review, test and publish a script