top of page
Search

SentinelOne Singularity Operations Center - Vulnerabilities, Risk Scores and Findings

4 days ago
12 min read

Updated: 7 minutes ago


Before starts



Now, vulnerabilities and application scanning. I have touched this before in the older console, and I am not going to write it all again. If you want the fundamentals, go here first:



Read that one and come back. This article is only about what changed.


And I will say it straight away, because in previous I was complaining the whole time: this one is actually good. There is a lot more here than the old console had, and most of it I like. So let us start with the thing that I think matters most.


You can finally run a scan when you want to

Remember the old console? You scheduled the scan. That was it. Wednesday, 10:30 AM PST, every week, and that was your lot. If someone asked you on Thursday afternoon whether a machine got patched, you either waited a week or you went and looked at the endpoint yourself.


Now there is an on-demand scan. You click Scan Now and it goes.

I know, it sounds small. It is not. Think about why vulnerability data was never actually useful before — it was always old and you could not refresh it. Now you push a patch, you scan, you know the same day. That is a real difference in how you work, not a feature bullet.


A few things to keep in mind:

  • You cannot run another manual scan for 8 hours. So do not plan to sit there re-scanning every twenty minutes watching a rollout. Run one scan at the end of your window, not five during it.

  • Scheduled and manual scans run in parallel, so the cooldown does not interfere with your weekly schedule.

  • After the scan finishes, the inventory still takes a few hours to populate. Scan complete does not mean data is there. Do not panic in the first hour.



Now where do you actually find this

This is the part that annoyed me. It is not on the Vulnerabilities page, which is where you will look first. It is here:


Policies and settings → Exposure Management → Application & OS Vulnerability

There is a Manage scan settings link from Inventory → Applications as well (S-25.2.6), and honestly that is how most people find it, by accident.


Three things on that page you need to know:

  • Scanning is off by default from Platform version X SP1. Both automated and manual. So if you are staring at an empty Vulnerabilities page wondering what is broken — nothing is broken, it was never switched on. Go and check yours right now, I will wait.

  • Settings are scope-based with inheritance. You will see Policy is inherited from a higher scope or Policy no longer inherited. Edit Policy breaks it, Inherit Policy gives it back.

  • The main toggle got renamed to Application & OS Vulnerability Management. It used to be Singularity Vulnerability Management. Same switch, do not go hunting for the old name.


And here is something worth stopping on, because it will catch you.

Inheritance here flows downward — parent to child, until a child refuses it.

In Previous console, the automatic alert action rules worked the opposite way: site beats account beats global. Same console, two features, two completely different mental models. Lovely.



On scheduling, since you can still do it. Before S-24.3.6 everyone got Wednesday 10:30 AM PST. Now the default is whenever you happened to click the toggle, and you set your own under Scan Execution Schedule — how often in weeks, which day, what time.

Set it properly. Otherwise you will discover you are scanning the whole estate at 2 PM on a Tuesday, and someone will ask you why.

Basic, Extensive, and the add-on — this is three things, not two

Right, this is where everyone gets confused, including me the first time. There are two scan types and one separate licence, and they are not the same conversation.


  • Basic Vulnerability Scan — you get this with Control or Complete. Windows and macOS applications and their vulnerabilities. From S-26.3.1 it also does macOS OS-level detection, but that bit needs the add-on.

  • Extensive Vulnerability Scan — needs Singularity Vulnerability Management. This is what gets you Windows and Linux OS-level detection, missing patches, extra intelligence. It runs the Basic scan too, so do not enable both.

  • Vulnerability Management add-on licence — different thing again. This controls what data comes back to you.


Now the one that really matters, and I want you to go and check this after you finish reading.

To detect Linux vulnerabilities at all, you must turn on the Extensive scan.

Basic does not do Linux applications or Linux CVEs. Nothing. So if your Linux fleet looks beautifully clean, I am sorry, but it is probably not clean — it has never been looked at.


Here is what the add-on actually buys you:

Data

Basic, add-on off

Basic, add-on on

Application CVEs, NVD data

Yes

Yes

Third-party intelligence

Only if Extensive ran

Yes

Improved CVE accuracy (Win 22.3+)

Only if Extensive ran

Yes

Linux vulnerabilities (Linux 23.3+)

Only if Extensive ran

Yes

OS CVEs

No

macOS yes; Win/Linux need Extensive

Temporal scores

No

Yes

Exploit maturity, in the wild, remediation level

No

Yes


Two things fall out of that table:

  • The SentinelOne Risk Score only exists with the add-on. No add-on, you get CVSS and nothing else. I will come back to why I think that matters more than it sounds.

  • The Extensive scan is heavy. Roughly 40MB on Windows, 45MB on Linux, pushed to every endpoint in scope and refreshed every one to four weeks, plus 3MB on first run. If you have a big estate on metered or bad links, have that conversation before you tick the box, not after someone notices the traffic. Agent floors are Windows 22.3+ and Linux 23.3+.


The Vulnerabilities page itself

Everything from every surface lands here — SentinelOne's own scanning plus partner findings like Wiz or Snyk.


Two things before you touch anything:

  • The date selector filters on last reported, which means last scanned. Not when the finding was found. It defaults to All time and I would leave it there until you know what you are doing, because narrowing it does not do what you think.

  • Show Filters opens the panel, but from S-25.1.2 you only see the common filters. Surface-specific ones, and anything with no matches, are hiding behind More filters. Click it once just so you know what you have.


Group by

  • You get up to four categories.

  • The order you pick them is the order they nest. Severity then Status gives you severity groups with status inside. Pick them the other way round and you get the opposite. This catches everybody once, including me.

  • Favourites sit at the top of the menu from S-26.2.1.

  • And the one people miss: you can bulk-action a whole group. See that Assigned To arrow on the group row? Use it and every finding in that group goes to one person in a single click. If you are splitting remediation work between teams, that is your mechanism right there.



Now this one I really like

From S-26.2.3 you can select any cell value in the table and get Exclude, Include as filter, or Copy.


Say one noisy CVE is dominating your list. Click it, click Exclude, done — the filter is built for you.

No filter panel, no hunting for the right field name, no typing out CVE IDs.


So instead of building your view forward through the filter panel, you build it by clicking things out of it. It is much faster, and if I were teaching someone this page on day one this is what I would show them first. The documentation barely mentions it and it is the feature I use most.



What happens to resolved findings

  • From S-25.1.2 anything confirmed resolved by a surface stays for 90 days. Before that version it just disappeared and you never saw it again.

  • Turn on Show Confirmed Resolved to see them. Manually resolved ones show up even with that toggle off.

  • S-25.1.3 added a nice touch: update the finding within those 90 days — even just the description — and the clock restarts.

  • But remember: Resolved and untouched for 90 days means gone.

  • One quirk that will confuse you. The UI does not tell you whether something was resolved by a surface or by a person. You work it out from what you can do to it — surface-resolved gives you a note only, manually resolved gives you note, assign, status and verdict.


Opening a CVE

The risk score, and the thing I want to argue about

This is where the add-on earns its money, and rather than give you theory, let me just show you. Two findings from an estate I look after, both live right now:


Finding A

Finding B

Software

Google Chrome

Linux kernel

Severity

Critical

High

NVD base score

9.6

7.8

EPSS score

not enriched

0.98

Exploit maturity

not enriched

High

Exploited in the wild

not enriched

True

Fix available

not listed

Yes, published

Asset

a macOS laptop

a Linux server

Now, sort that by CVSS like almost everyone does out of habit. A sits on top. 9.6 beats 7.8, Critical beats High, job done, go home.


But look at what is actually true. B is being exploited in the wild right now. It has a 0.98 EPSS score, which is roughly a 98% modelled chance of exploitation attempts within 30 days. It has a mature public exploit. It is on a server. And there is a patch available this very moment. Meanwhile A is a browser bug, on a laptop, in a product that updates itself, with no evidence anyone is exploiting it.


B belongs at the top of your queue. CVSS puts it second.

That gap is the whole argument for the risk score, and it is why, if you ask me, the add-on is worth having rather than the line item you cut when budget gets tight. The risk score weighs active exploitation, EPSS probability, whether a fix exists, and threat intelligence — and it keeps recalculating. CVSS is a number someone worked out once and never looked at again.


One honest thing, and you can see it in my own table. Enrichment is uneven. A has no EPSS, no maturity, no in-the-wild flag at all. Some of that is the licence split I showed you above, and some of it is just that a CVE published four days ago has not been EPSS-scored yet. So please do not read a blank EPSS column as "safe". It means no data, which is a different thing entirely.

Two more from the documentation you should carry around:

  • EPSS and KEV data apply only to SVM surfaces.

  • For Linux assets, severity comes from the distribution security advisory, not NVD. So a Linux severity and a Windows severity are not the same measurement. Do not compare them like they are.


Evidence, timeline, indicators

Evidence gives you the CVE ID, NVD base score, the EPSS score and percentile (S-24.3.3), and the description.


Keep those two EPSS numbers straight, because I hear people use them interchangeably and they are not the same:


  • Score — the modelled probability of exploitation in the next 30 days.

  • Percentile — where that score ranks against every other scored CVE. 90% percentile means it is more likely to be exploited than 90% of everything else scored.


For prioritising, I find the percentile more useful, because it tells you where to draw your line.


CVE Timeline draws the lifecycle for you — identified, CVE assigned, advisory published, mitigation published, PoC published, exploited in the wild, countermeasures, CISA KEV entry added. Look at the gap between PoC published and exploited in the wild. That gap is the window you are actually racing, and I will tell you from experience, showing management that picture works better than any number you can quote them.


Risk Indicators come grouped Critical down to Very Low, top three per group up to ten, with Show more for the rest.

There are about fifteen of them in total. I am not going to list them all out, because you will not read them. These are the three I actually filter on:

  • Exploited in the wild — it is real, not theoretical.

  • Patch available — you can actually finish the job.

  • Does not require user privileges — it is cheap to attack.


Below that sit the CVSS vector components and the temporal values, exploit maturity and reported confidence. SentinelOne is on CVSS v3.1 at the moment.



Remediation

With the add-on you get Remediation Insights — Linux agents 23.3+, Windows agents 24.2.2+, on by default from 24.4.1.

There is a Knowledge Base link out to the vendor page, which sometimes has the patch sitting on it.

And this, for me, is the practical heart of the whole page: the Remediation Insights Available filter together with the Fix Version column.

Fix Version tells you the exact version that closes the finding. Filter down to the things that have one, and suddenly you are not looking at a list of problems — you are looking at a list of work that can actually be finished. That is a much better thing to hand someone.


The rest of the tabs, quickly

  • Other affected assets — everything else carrying the same CVE, with CSV export.

  • Notes — view and add, nothing surprising.

  • History — every action and when it happened. This is where you go when a verdict looks wrong and you want to know who did it.

  • Graph — the asset-to-finding relationship, for scope and impact.



Statuses and verdicts

  • Statuses: New, In Progress, Resolved, Risk Acknowledged.

  • S-25.1.5 merged On Hold and To be Patched into In Progress. Existing findings keep the old values and you can still filter on them, but you cannot set them on anything new. So when you see statuses in your data that you cannot apply anywhere — that is why, it is not a bug.

  • Risk Acknowledged is the one I would use properly. It is the honest option for "we know about it, we have decided not to fix it, here is why". Put a note with it and you have proper record-keeping, which is much better than quietly marking something Resolved and hoping nobody asks.

  • Verdicts: True Positive, False Positive, Undefined. Everything starts Undefined and unreviewed sorts first, which for once is the right default.

  • From S-25.1.3 you can change status, verdict and assignment straight from the table without opening anything. On a page with this many rows, that saves you real time.



Event Search

Click Event Search on a finding and it queries that one. But more usefully, you can query the whole class.

Select the XDR view and use these:

class_uid = '2002'            vulnerabilities
class_uid = '2003'            misconfigurations
class_uid in ('2002','2003')  both

Those are OCSF class IDs, and they are your bridge between this page and everything else you do. If you are building any custom reporting or correlation off vulnerability data, start there.



Getting reports out


Reports, and there are two formats:

  • CSV (S-25.3.6) — up to 1 million rows for vulnerabilities and misconfigurations, 100,000 for alerts.

  • PDF of grouped views (S-26.1.1) — each group row becomes a section. Up to 1,000 records, and you must pick exactly one Group by value.

  • Same flow for both: set your view, scope, tab and filters, then Export → download now, or Schedule CSV / PDF for recurring email.

  • The permission moved at S-25.4.2 to Custom Reports. It used to be Reports, so if someone suddenly cannot get to their reports, that is your answer.

  • From a group scope you can only download. No scheduling, no email. Go up to site or account.


The grouped PDF is what I would use for client reporting. Group by asset category or software name, export it, and you get something organised the way a client actually reads — rather than a CSV they will open once and never again.


Application Inventory over at Inventory → Applications works the way it always did — every detected application whether it is risky or not, same Scan Now, same 8-hour cooldown. One thing to know: Macs on agents older than 22.2 GA will not show up there at all.


Now let me show you what this looks like for real

Here is the thing nobody writes down. Documentation tells you what the buttons do.


It never tells you what the page looks like when you actually have an estate behind it.

The one I am working in has 1,831,548 findings sitting in it right now. 1,506,234 application, 325,316 OS. And if you filter down to Critical only, you still have 110,796 of them.


A hundred and ten thousand criticals is not a work queue. It is a reason to close the tab and go and do something else.


So here is the funnel I would actually build, and the order matters:

1. Severity: Critical and High. Just a starting point, nothing clever.

2. Exploited in the Wild: True. This single filter does more than all the others put together. It is the line between theoretical and real.

3. Remediation Insights Available, or a Fix Version that is not empty. If there is no fix, that is not remediation work — that is risk acceptance work, which is a different queue and a different conversation with a different person.

4. Asset Category: Server for the first pass. Workstations after.

5. Group by Software Name. Watch thousands of rows collapse into a handful of upgrades, because one Chrome update closes hundreds of them at once.

6. Assign the whole group with that Assigned To arrow I showed you.



And that is how an unusable number becomes a list of software upgrades with somebody's name on each one. Filter on exploitation before you rank by severity, and group before you assign. Do it the other way round and you will be there all day.


So what do I think

Older console Vulnerbiltiy management I did not like, and I said so. This one I do.

My complaints, and I do have them:

  • Scanning being off by default is going to catch people out.

  • Scan settings are buried where nobody would think to look, the Basic versus Extensive versus add-on thing takes three reads before it sticks.

  • Linux quietly producing nothing until you enable the Extensive scan is the kind of default that gets an entire fleet missed for a year.


So go and check that toggle. And then go and check whether your Linux estate has ever actually been scanned. I think some of you are going to find out it has not.


See you in the next article.

I will meet you in the next article. Till then, keep learning.



                                                                    - written from hands-on work with the console, By Dean



--------------------------------------------------Cyberengage--------------------------------------


 
 
 

1 Comment


Markus White
3 days ago
Edited
Like

Ready to discuss:

- Schedule a call for a consultation

- Message me via "Let's Chat" for quick questions

​

Let's connect!

Subscribe to our newsletter

Connect With Me:

  • LinkedIn
  • Medium

© 2023 by Cyberengage. All rights reserved.

bottom of page