Skip to content

Correlations

Correlations maps what your events connect: which identities share an address, which address reached many accounts, which applications a user touched. Use it to answer “who else is affected?”

The page heading reads Correlations, with the line “Map the identities, addresses and apps your events connect, and work the findings that matter first.”

Where Correlations under Security & monitoring in the sidebar
Who can use it Admins, members and viewers
Plan Every plan. No feature switch is needed.

For step-by-step tasks, see Run a correlation.

The page has four parts, from top to bottom:

  1. Correlation Explorer: the controls that decide which events are loaded.
  2. Findings, with the entity panel under it, on the left.
  3. Entity graph on the right.
  4. Event Timeline at the bottom.

A correlation is a query. It loads a sample of events, then builds the findings, the graph and the timeline from that sample only.

The query runs:

  • once when you open the page,
  • again every time you change Scope,
  • when you choose Run Correlation.

The controls inside the Entity graph card (the slider and the tick boxes) only change what is drawn. They never load new events.

The card is titled Correlation Explorer, with the line “Inspect related signals within a scope to identify coordinated attack patterns.”

Control What it does
Scope Chooses the tenant or correlation group to query. The placeholder is Select scope.
Time Range Two date and time pickers: Start Date & Time and End Date & Time. The default is the last 24 hours, in your browser’s local time. The start cannot be later than the end.
Event Sample Size 100 events, 200 events, 500 events (default) or 1000 events. The helper text reads “Higher limits improve coverage but increase correlation runtime.”
Filters Optional rows that narrow the events. See Filters.
Run Correlation Runs the query. It shows a spinner while it loads. It is disabled while loading, and when the workspace has no tenants.
Reset Removes all filter rows, deselects the node, clears the timeline filter and clears the focus. It does not change Scope, Time Range or Event Sample Size, and it does not run the query again.
  • Correlation groups come first, in alphabetical order. A group reads <Group name> (group, N tenants).
  • Tenants that are not in a group follow, by tenant name.
  • A tenant that belongs to a group is not offered on its own. Choose its group instead.
  • The default scope is the first one in the list. If the workspace has a group, that is a group.
  • A group scope queries at most 50 tenants.
  • In a group, workforce tenants (Okta and Entra ID) and customer tenants (Auth0) appear in the same graph, but a workforce identity is never joined to a customer identity.

An admin creates correlation groups under Incident settings. See also Correlate Okta and Entra ID.

The query returns the most recent events in the time range, up to the size you choose. If more events exist, the older ones are left out. The findings, the graph and the timeline all use this sample only.

For a busy tenant, a long time range with a small sample shows only the end of the range. Shorten the range, add filters, or raise the sample size.

Choose Add filter to add a row. With no rows the card shows “No filters applied. Use “Add filter” to refine the correlation scope.”

Each row has a field, an operator, a Value box and a bin icon (Remove filter).

Part Choices
Field Event Type, Severity, Actor Display Name, Actor Alternate ID, Target Display Name, Client IP
Operator Equals, Contains, Not equals
Value Free text, up to 256 characters

Rules:

  • Matching ignores upper and lower case.
  • A row with an empty value is ignored.
  • Use one row for each field. If two rows use the same field, only the last one applies.
  • You can apply at most 20 filters.

Actor Alternate ID is usually the sign-in name, such as an email address. Actor Display Name is the name shown for the person.

Two boxes can appear above the results.

Banner When it appears What it means
Retention window notice The time range starts before the workspace’s payload retention cutoff, or some events have a redacted payload The events are still included. The client IP and the raw payload can be missing for the older ones. The text gives the counts and the cutoff, for example “3 events older than 48h” and “This affects payload detail visibility, not event inclusion.”
“⚠️ Showing up to N most recent events. Apply filters or shorten the time range for more focused results.” The result reached the Event Sample Size More events exist in the range than were loaded. Older events are not in the result.

The date pickers accept any range. Nothing stops you from choosing a start date before the retention period. Events older than the retention period no longer exist, so they cannot be found.

The card is titled Findings, with the line “Deterministic patterns in this scope and window. Select one to focus the graph.” Findings are fixed rules. The same events always give the same findings.

Each row shows:

  • a severity badge: high, medium or low,
  • the number of events and the time they span, such as “14 events · 12m”. The span can read instant, or a number of minutes, hours or days,
  • the title,
  • one sentence that explains why the pattern matters,
  • Next step: with the recommended action.

Findings are sorted with the most severe first, then by event count. The list shows at most 25. With none, the card reads “No findings in this window.”

Select a finding to work on it:

  • The graph highlights the finding’s nodes and their direct neighbours. Everything else is dimmed.
  • The Event Timeline shows only the finding’s evidence events, with “Filtered to <finding title>.” and a Clear button.
  • Select the same finding again to deselect it.

The timeline lists at most 200 evidence events for one finding. The event count on the finding is always the true total.

In the titles below, <N>, <F> and <C> are numbers.

Title <ip> used by <N> identities, with , <F> failures when there are failures
Appears when One IP address is used by 3 or more identities
high 5 or more failures, across 3 or more identities that failed
low Every event succeeded
medium Everything else
Why “Several identities acting from one address is how password spraying and shared-credential abuse look; an office NAT or VPN looks the same but succeeds.”
Next step “Check whether this address is a known egress point; if not, review the failing identities for lockouts and block the address.”
Title <identity> seen from <N> IPs, with in <C> countries when a country is known
Appears when One identity is seen from 3 or more IP addresses, or from 2 or more countries
high 2 or more countries and at least one failure
medium Everything else
Why “One identity across many addresses or countries in a short window can mean a stolen session or credential in use elsewhere.”
Next step “Confirm travel or VPN use with the user; if unexplained, revoke sessions and reset credentials.”
Title <identity>: <N> failures, then a success
Appears when 5 or more failures in a row are followed by a success within 1 hour of the last failure
high The success came from an IP address not seen earlier for that identity in the sample
medium Everything else
Why “A run of failures ending in a success is the signature of a guessed or stuffed password.”
Next step “Verify the successful sign-in with the user, revoke the session if it was not them, and enforce MFA.”

MFA changed, then a sign-in from a new address

Section titled “MFA changed, then a sign-in from a new address”
Title <identity>: MFA changed, then a sign-in from new IP <ip>. The title ends with (no earlier activity in window) when the sample has no earlier address to compare with.
Appears when An MFA change is followed by a sign-in from an IP address not seen before. MFA events within 60 seconds of each other count as one change.
high The address is new compared with the identity’s earlier addresses in the sample
medium There was no earlier activity to compare with
Why “Changing MFA and then signing in from an address never seen before is the classic account-takeover sequence.”
Next step “Confirm the MFA change was requested; if not, suspend the account, revoke sessions and re-enrol factors.”
Title <identity> made <N> privilege changes, or An unattributed actor made <N> privilege changes when the events name no actor
Appears when The sample holds any event that grants privileges or sets up persistence, such as a role, group or credential grant
high Any of the events has severity CRITICAL or ERROR
medium Any of the events has severity WARN
low Everything else
Why “Role, group and credential grants widen what an identity can reach and are a common persistence step.”
Next step “Check each change against an approved change request and remove any that are not.”
Title <identity> is active in <N> tenants
Appears when The scope is a group with more than one tenant, and the identity is seen in 2 or more of them. It never appears for a single-tenant scope.
medium The identity has failures in 2 or more tenants
low Everything else
Why “The same identity active in several tenants of a group ties their activity together, which matters when one of them is under attack.”
Next step “Review this identity in each tenant; failures in more than one suggest a campaign rather than a typo.”
Title <identity>: <N> provider-flagged events, or <tenant name> (no identity): <N> provider-flagged events
Appears when The sample holds events that the identity provider marked as a risk signal, or events with severity CRITICAL or ERROR
Severity Always high
Why “The identity provider itself rated these events as risky or high severity.”
Next step “Open the evidence events and follow the provider risk guidance; confirm or dismiss the risk in the provider console.”

For how severities are set, see Severity.

A Related incidents box appears at the top of the Findings card when open or escalated incidents contain events from the sample.

  • It lists up to 10 incidents, highest score first.
  • Each line is a link with the incident’s entity, then its severity, status and event count.
  • Select a line to open the incident.

The box only appears when automatic incidents are on for the workspace. See Incidents.

This is the only link between the two pages, and it goes from Correlations to the incident. An incident page has no link to Correlations. To go the other way, see Start from an incident.

The Entity graph card draws the sample as nodes and links. A node is an identity, an IP address or a target such as an application. A link between two nodes means they share events.

The page has no legend. Use this table.

What you see Meaning
Circle An identity
Rounded square An IP address
Diamond A target, such as an application
A larger node The node has more connections
Red ring around a node The node is part of a high finding
Amber ring around a node The node is part of a medium finding. A low finding gives no ring.
A thicker link The two nodes share more events
Red dashed link At least half of the events on that link are failures
Grey solid link Fewer than half are failures

Labels are shown only for the selected node, the node under the pointer, nodes in a high or medium finding, and highlighted nodes. Labels longer than 24 characters are cut short.

Hover for details:

  • On a node: the label, its number of events and its number of connections.
  • On a link: the number of events, the number of failures and the most common event types.
Control What it does
Counter Reads “<n> nodes, <m> links”.
Identities, IPs, Targets Tick boxes, all on by default. Untick one to hide that type of node.
Minimum shared events A slider from 1 to the largest number of events on any link. Raise it to hide weak links. It returns to 1 with every new result.
Failures only Keeps only links that carry failed activity.
Show common apps Shows targets that most identities use. They are hidden by default. A target is common when the sample has at least 5 identities and half or more of them touch it.
Clear focus Appears when a node or a finding is highlighted. It returns the graph to the full view.
Zoom out, Zoom in, Reset zoom Three icon buttons at the top right of the drawing.

A node with no link left after these filters disappears from the graph. If nothing is left, the card reads “Nothing to draw with these filters.”

The graph draws at most 150 nodes. Nodes that appear in findings are kept first, then the nodes with the most connections, then those with the most events. When nodes are left out, the counter adds “<k> lower-activity nodes not shown”.

Action Result
Select a node Opens the entity panel for it
Select the empty background or a link Deselects the node
Escape Deselects the node
Tab, then Enter or Space Moves to each node in turn and selects it
Drag a node Moves the node
Scroll or pinch Zooms
Drag the background Pans

Focus dims the part of the graph that you are not working on.

  • Focus on the entity panel dims everything further than 2 steps from the selected node.
  • Selecting a finding dims everything further than 1 step from the finding’s nodes.
  • Clear focus removes either one.

If the graph filters hide the focused node, the card reads “The focused node is hidden by the current filters.” with its own Clear focus button.

Focus also writes the scope and the focused node into the browser address. Copy the address to share the view with another member of the workspace. The time range, sample size and filters are not in the address. The link opens with the default last 24 hours. If the focused node is not in the new result, the page opens without a focus and shows no message.

Select a node to open its panel under Findings.

Field What it shows
Header line The node type (Identity, IP address or Target) and the domain (Workforce or Customer), then the node’s label
Events Number of events in the sample that involve the node
Failures How many of those events failed
Connections Number of nodes it is linked to
First seen, Last seen The first and last event times in the sample
Countries Countries seen for the node. Shown only when there are any.
Top event types The most common event types
In findings Titles of the findings that include this node. Shown only when there are any.

Buttons:

  • Focus: see Focus.
  • Show events: filters the Event Timeline to this node’s events.
  • The X (Close entity panel) closes the panel and deselects the node.

The card is titled Event Timeline, with the line “Detailed view of all correlated events in chronological order”. It lists the events in the sample.

  • The line “Showing <n> of <total> events.” gives the count.
  • When a finding or Show events is active, it adds “Filtered to <label>.” and a Clear button.
  • A line that starts “Retention:” can appear when some events are older than the payload retention cutoff. It means the client IP and raw payload can be missing for those events. The events themselves are still listed.
Column What it shows
Event The event type and message
Actor Who did it
Target What it was done to
Client IP The source address
Timestamp When it happened
Severity A coloured severity chip
Inspect A button that opens the event below the table

The table shows 25 rows on a page. You cannot sort or export it.

Inspect opens an Event focus card under the table, with the line “Context for the highlighted event.” It shows Event type, Timestamp, Actor, Client IP, Message, Targets and Raw payload.

In Event Details, on the dashboard feed or the Events page, choose Explore correlations. The button appears for any event that has a tenant. It sits beside View incident when the event is also part of an incident.

The page opens like this:

  1. Scope is set to the tenant’s correlation group, or to the tenant when it is in no group.
  2. The query runs for the default last 24 hours with 500 events.
  3. If the event is in the result, the graph is focused on the event’s identity and the event’s row is highlighted in the Event Timeline. The timeline opens on the page that holds the row.

There is no Explore correlations button on an incident page or in incident evidence.

Older Slack alerts, ServiceNow tickets and bookmarks can hold a link that ends in /events/correlate. These links now open the Incidents page. If the link names an event that is part of an incident, that incident opens. If not, you see “That event is not part of an incident”. Open Correlations from the sidebar instead.

Auth0 can hide personal data in its log stream. Identities then arrive as hashed values. See Auth0.

For a masked identity, Correlations skips two findings:

  • Identity seen from many addresses.
  • Identity active in several tenants.

The other findings still work. The graph shows the hashed value as the identity’s label.

The Correlations page has no Ask Alex AI button. You can still open Ask Alex from the sidebar. Its header reads “Viewing: Correlation graph and findings”.

Most of these appear as a short pop-up message.

What you see Why What to do
“No tenants are configured for this workspace yet. Add a tenant to start exploring correlations.” Scope and Run Correlation are disabled. The workspace has no tenants. Connect a tenant. See Quickstart.
“Run a correlation query to visualise related activity across the tenant.” No result is loaded. The first query failed, or no scope is selected. Choose a Scope and choose Run Correlation.
“You have no events to correlate yet.” The query found no events to work with, or the scope could not be read. Check that the tenant is sending events on the dashboard feed and on Tenant health. Widen Time Range and run again.
“Could not load correlation data.” A network problem, or an error with no detail. Check your connection and choose Run Correlation again.
“Correlation query failed (<status>)” The server returned an error with no text. Run it again. Status 402 means the workspace is paused for billing. Status 403 means you no longer have access to the workspace.
“Unknown correlation scope” The group you selected no longer exists. It was changed after the page loaded, or you followed an old shared link. Reload the page and choose a scope again.
“This group has more than 50 tenants” A group scope can query at most 50 tenants. An admin splits the group under Incident settings.
“Invalid time range provided” The start or end date is empty or cannot be read. Enter both dates again.
“At most 20 filters may be applied at once” There are more than 20 filter rows. Remove filters.
“Filter values may be at most 256 characters” A filter value is too long. Shorten the value.
A message that starts “Unsupported filter field; use one of”, or “Unsupported filter operator; use equals, contains or not_equals” The page in your browser is older than the current version of the product. Reload the page fully.
“Correlation query too large” The request is too big. Remove filters.
“Failed to compute correlations” An internal error. Run it again. If it continues, lower Event Sample Size or shorten Time Range, then report it to the product team.
Retention window notice The range reaches past the retention cutoff, or some events have a redacted payload. This is expected. Client IP and raw payload can be blank for older events. An admin can raise retention under Policies if the plan allows it. The change applies to new events.
“⚠️ Showing up to N most recent events. Apply filters or shorten the time range for more focused results.” The result reached the sample size. Shorten the range, add filters, or raise Event Sample Size. The maximum is 1000 events.
A filter seems to be ignored Two rows use the same field, so only the last one applies. Or the Value box is empty. Or you did not run the query again. Keep one row for each field, fill in the value, then choose Run Correlation.
A tenant is missing from Scope The tenant belongs to a correlation group. Choose the group.
What you see Why What to do
“Nothing to draw with these filters.” The graph controls removed every link. Lower Minimum shared events, untick Failures only, tick Identities, IPs and Targets, or tick Show common apps.
“The focused node is hidden by the current filters.” The graph controls hide the node you focused on. Choose Clear focus, or relax the graph controls.
“<k> lower-activity nodes not shown” The graph draws at most 150 nodes. Shorten Time Range or add filters, then run again.
“No findings in this window.” No pattern reached its threshold in the sample. This is not an error. Widen the range or raise the sample size if you expected one.
No “is active in N tenants” finding The scope is a single tenant, or the identity is masked by Auth0. Choose a group scope.
No Related incidents box No open or escalated incident contains events from the sample, or automatic incidents are off for the workspace. Nothing to do. See Incidents.
“No events matched the correlation query.” and “Try widening the time range or adjusting filters to load events.” The query returned no events. Widen Time Range or remove filters, then run again.
“No events match the current filter.” and “Clear the filter to see every event in this window.” The finding or node filter on the timeline matches nothing. Choose Clear.
Client IP or Raw payload is empty for older events The payload detail was removed after the retention cutoff. Expected. See the Retention window notice.
What you see Why What to do
You chose Explore correlations but nothing is highlighted, with no message The event is older than 24 hours, or is not among the 500 most recent events of the scope. Set Time Range around the event, raise Event Sample Size and choose Run Correlation.
A shared link opens without a focus The focused node is not in the last 24 hours of events. The link does not carry the time range or the filters. Ask the sender for the time range and filters, set them, and run again.
“That event is not part of an incident” You followed an old “correlate” link from Slack or ServiceNow. It now opens Incidents, and the event has no incident. Open Correlations from the sidebar and run a query for the event’s tenant and time.
“Could not check whether that event belongs to an incident.” The lookup failed after an old link. Reload the page.

For other problems, see Troubleshooting.