SentinelOne Singularity Operations Center - Graph Explorer: Seeing Your Detections as Relationships Instead of Rows

Before starts
Check Out old Series > https://www.cyberengage.org/courses-1/mastering-sentinelone%3A-a-comprehensive-guide-to-deep-visibility%2C-threat-hunting%2C-and-advanced-querying%22
Every other screen we have covered in this series gives you rows.
The alerts list is rows.
Deep Visibility is rows.
The vulnerability page is rows.
Rows are great when you already know what you are looking for.
Graph Explorer is the one screen in the Ops Center that does not give you rows. It gives you dots and lines. Every dot is a thing (an endpoint, a user, a cloud resource, an alert, a vulnerability, a misconfiguration) and every line is a relationship between two of those things.

So instead of asking "show me the alerts", you ask "show me the machines that have alerts", or "show me the machines that have a critical alert and also a critical vulnerability". That second question is very hard to answer from two separate lists. In the graph it is one query.
What it is, and what it is not
Graph Explorer is part of the Singularity Graph, and it is only available once the Operations Center is enabled. You will find it in the sidebar as Graph Explorer.
The first thing to understand is how it differs from the Storyline, because people mix the two up.
Storyline correlates events on one endpoint. Process started, process wrote a file, file made a network connection. It is the story of what happened on that machine.

Singularity Graph correlates assets, alerts and exposures across your environment based on how they relate to each other. It is the map of what is connected to what.
When you click an alert node in the graph, the alert details open, and from there you can still jump into its Storyline.

So the two work together. The graph tells you how wide the problem is. The Storyline tells you what actually happened.
The second thing to understand is where the data comes from. The graph supports every asset type that is in the Unified Asset Inventory. That means endpoints, discovered devices, identity assets and cloud assets. But it only knows about what your tenant actually feeds it.
If you only run the EDR agent, your graph will mostly be endpoints and the alerts on them. If you also have identity, cloud security or network discovery feeding the inventory, the graph gets much richer, because there are more things to connect. Keep that in mind before you judge the feature on a small tenant.
The pieces of the screen
There are four parts to learn.
The query builder. This is where you build a query. You pick a starting resource, like Windows desktop or macOS desktop, then click the plus sign to add either a property (a filter on that thing, like Asset status = Active) or a relationship (a link to another thing, like Alerts - has detection of).


One rule matters here: a filter applies to the item directly above it, or to the item it is connected to by a line in the builder.

So if you add a Severity filter right under an Alerts step, it filters the alerts, not the machines.
The builder only offers relationships that make sense for the step you are on. If you are on an endpoint, it will offer alerts, vulnerabilities, networks and so on. You do not need to memorise anything. Click the plus sign and read the menu.
The graph itself. The query runs as you build it, and the graph redraws each time. You get the first 20 nodes. If there are more, a Partial results note shows and you click Load more under the graph. The product limits page puts the maximum number of assets returned at 100, so this is not a tool for listing your whole estate. It is a tool for looking at a slice of it.

The controls. Along the side of the graph you can switch between graph view and table view, change the layout, open the key that explains each node icon, re-centre, zoom, and export. There is also a filter option called Auto Grouping, which groups assets together when five or more are linked to the same node. Leave that on. Without it, a busy graph turns into a ball of string.

The Library. This is where queries live. It has four categories:
Recently Used - the last 20 queries you ran
Suggested Searches - ready-made queries from SentinelOne
My Searches - queries you saved as Private
Public Searches - queries other people in your scopes saved as Public
Five queries anyone can build
These all use plain resource types that exist in any tenant with agents in it. The exact wording in your menus may be slightly different depending on your console version, but the shape will be the same.
Before you start, pick one site in the scope selector. The graph is much easier to read on one site than across a whole account, and you will hit the 100-asset ceiling much later.
Query 1: Which machines have alerts
This is the "hello world" of Graph Explorer.
1. Open Graph Explorer.
2. In Select resource, type Windows and pick Windows desktop. (Use macOS desktop if that is what your estate is mostly made of.)
3. Click the plus sign and pick Alerts → Alerts - has detection of.

That is the whole query. You get every Windows desktop in scope that has at least one alert, with each alert drawn as its own node and a line to the machine it fired on.
What to look for: a single alert node with lines to many machines. That is one detection firing across your estate. Sometimes that is a real campaign. More often it is one noisy rule. Either way, you see it in a second instead of scrolling through a list.
Query 2: Only the active machines
Same as Query 1, with one property added.
1. Start with Windows desktop.
2. Click the plus sign, search asset status, and select Asset status.
3. Equals is selected by default. Pick Active and click Add, then Add again to put it into the query.
4. Click the plus sign and add Alerts → Alerts - has detection of.

This is how you clean up the graph. Machines that have not been seen for weeks are not your priority, and this one property removes them.
Query 3: Only the serious alerts
Now filter the alerts instead of the machines.
1. Build Query 2.
2. On the Alerts step, click the plus sign and add a Severity property set to Critical (and High, if you want both).

Remember the rule from earlier. Because the filter sits under the Alerts step, it filters the alerts. The machines that only had low-severity alerts drop out of the graph, because they no longer have a matching relationship.
Query 4: Machines with an alert and a vulnerability
This is the one that shows why the graph exists.
1. Start with Windows desktop, Asset status Active.
2. Add Alerts → Alerts - has detection of, Severity Critical.
3. Go back to the Windows desktop step, click its plus sign, and add Vulnerabilities → has detection of, with Severity Critical.

Now you only get machines that have both: a critical alert and a critical vulnerability. That is a very short list, and it is the list I would want on my screen first thing in the morning. A machine that is being attacked and also has a known hole in it is the machine to look at before anything else.
Try answering the same question from the alerts page and the vulnerabilities page separately and you will see why I like this query.
Query 5: Where does this one vulnerability live
This one starts from the other end.
1. In the sidebar, open Vulnerabilities and click any vulnerability name.
2. In the details, open the Graph tab.

You get the vulnerability in the middle and every asset it was found on around it. The default query behind it is: find a vulnerability with this CVE ID, a status that is not Resolved, detected on all discovered assets. Misconfigurations work the same way from the Misconfigurations page.

Click View in Graph Explorer and you will see the query that produced the picture. Change it, for example by adding Asset status = Active, and you have your own query built from a SentinelOne one.
The Graph tab inside every detection
This is the half of the feature you will probably use more, because you do not have to build anything.
Open any alert in the Ops Center and you will see a Graph tab next to the other tabs. It is already filled in. For an alert, it shows the related processes, the endpoints involved, the indicators, and other alerts connected to them. For an asset, it shows directly related assets, alerts and exposures. For a misconfiguration, it shows the misconfiguration and any AD users with it. For a vulnerability, where it was detected.

The graph is interactive. Click any node and you get two buttons:
Show entity info - opens the details of that node without leaving the graph
Expand more direct relationships - adds the next ring of connections. The options depend on what you clicked: related alerts, related assets, related misconfigurations, related vulnerabilities.

This is how I would use it during triage. Open the alert, open the Graph tab, click the endpoint node, and expand related alerts.

If nothing else appears, the problem is probably contained to that one machine and that one detection. If three other alerts appear on the same machine, or the same alert appears on other machines, the picture just changed and you know it before you have opened a single log.
Then click View in Graph Explorer to take that same picture into the full explorer, where you can edit the query, export it, or save it.
Saving, sharing and exporting
Once a query is useful, save it. With the query open, click Save As, then choose Private (only you) or Public (adds it to the Library for everyone in your scopes), and click Save. To open it again, click Load next to Save As.
If you work in a team. Save it as Public with a clear name and everybody on shift has the same morning list without having to rebuild it.
For exports, it depends on the view. In graph view, Export gives you an image (JPEG, PNG or SVG; the misconfiguration page also lists PDF). In table view, Export gives you a CSV of the results.
The image export is underrated. If you need to explain to a client or a manager why one alert matters more than another, a picture of one machine with a critical alert and two critical vulnerabilities hanging off it does that better than any table.
Limits and permissions in one place
First 20 nodes per run, then Load more.
Maximum 100 assets returned per query.
Recently Used keeps your last 20 queries.
Auto Grouping kicks in at 5 or more assets linked to one node.
Graph Explorer sits under the Unified Asset Inventory permission in Ops Center RBAC. As I read the table, Admin has full access and the other roles (C-Level, IR Team, IT, SOC, Viewer) can view.
It only shows asset types your tenant feeds into the Unified Asset Inventory.
Platform versions: the graph and Library go back to Z SP5, the Graph tab and saving queries came in S-25.1.1, and the current query builder is documented from S-25.1.3.
What I would do on Monday morning
Open Library → Suggested Searches and run three of them on one site.
Build Query 1 on that site and look for any single alert node linked to many machines.
Build Query 4 and save it as Public with a name your team will recognise.
Next time you triage an alert, open its Graph tab before you open the Storyline, and expand related alerts on the endpoint node.
So what do I think
I think the graph is honest about what it is.
It is not a replacement for Deep Visibility, and it will not tell you which command a user ran or which URL a file came from.
If you go in expecting that, you will be disappointed.
What it does well is the question that sits before all of that: how big is this?
Which machines
which users
what else is wrong with them.
That is the question you normally answer by opening four tabs and comparing lists in your head. Here it is one picture.
I will meet you in the next article. Till then, keep learning.
- written from hands-on work with the console, By Dean
--------------------------------------------------Cyberengage--------------------------------------


Comments