Use Cases
These examples show how to combine trigger conditions, step types, wires, placeholders and variables into complete flows in UTMStack v12. Each one uses a correlation rule that ships with UTMStack v12.0.2. Adapt the commands to your machines, then test every flow with Trigger flow on a test machine before you switch it on.
Flows act for real. Every example below changes machines, accounts or network rules. Read Building a Flow and Placeholders first, and never switch a flow on before you have tested it by hand.
To use an example so that it stays switched off while you test it:
Click New flow > From an existing one and pick any flow. Copies are saved switched off.
Open the Code tab and replace its whole content with the example YAML.
Adjust host names, lists and commands, then click Save.
Test with Trigger flow, check Execution history, and only then switch the flow on.
If you start with New flow > From scratch instead, the flow is switched on as soon as you save it.
Use case 1: Contain an SSH brute-force attack on a Linux server
Scenario. Someone guesses passwords over SSH on a Linux server. You want to block the attacker's address on that server, end the sessions of the account that was used, and tell the team. If the block fails, the team must know at once.
Trigger. Alert name is System Linux: Possible Brute Force Attack (shipped rule: ten or more failed SSH sign-ins on the same server within 15 minutes).
Steps.
| Step | Type | Runs | What it does |
|---|---|---|---|
step_0 | Shell (bash) on the alert source | First | Blocks the attacker's address. |
step_1 | Shell (bash) | When step_0 succeeds | Ends all processes of the target account. |
step_2 | Send notification (WARNING) | When step_0 succeeds | Tells the team what was blocked. |
step_3 | Send notification (ERROR) | When step_0 fails | Asks the team to act by hand. |
- name: Contain Linux SSH brute force
description: Block the attacker on the server, end the target account's sessions, notify the team.
conditions:
- operator: IS
field: name
value: 'System Linux: Possible Brute Force Attack'
roots:
- step_0
nodes:
step_0:
kind: executor
executor: shell
platform: linux
shell: bash
command: iptables -I INPUT -s $(alert.adversary.ip) -j DROP
onSuccess:
- step_1
- step_2
onError:
- step_3
step_1:
kind: executor
executor: shell
platform: linux
shell: bash
command: pkill -KILL -u $(alert.target.user)
step_2:
kind: executor
executor: notify
params:
type: WARNING
message: SSH brute force on $(alert.dataSource) - blocked $(alert.adversary.ip)
step_3:
kind: executor
executor: notify
params:
type: ERROR
message: Could not block $(alert.adversary.ip) on $(alert.dataSource) - act by handWhat happens. The agent receives iptables -I INPUT -s '203.0.113.45' -j DROP (UTMStack adds the quotes). On success, the account's processes are ended and a warning appears in the bell menu, reading for example SSH brute force on 'web-01' - blocked '203.0.113.45'. If the block fails, an error notification appears instead, and step_1 and step_2 show as Unreachable in Execution history.
Notes.
step_1andstep_2both hang off the success wire, so they run in parallel.If the alert has no target user,
$(alert.target.user)is sent as written andstep_1fails. Check a real alert of this rule first.An
iptablesrule added this way is lost when the server restarts unless you save it.
Use case 2: Stop an account takeover on Windows and open an incident
Scenario. A Windows machine records many failed sign-ins for an account, then a successful one from the same address. You want to block the address, disable the account and open an incident for the team to investigate.
Trigger. Alert name is Windows: Multiple Logon Failure Followed by Logon Success (shipped rule).
Steps.
| Step | Type | Runs | What it does |
|---|---|---|---|
step_0 | Shell (PowerShell) on the alert source | First | Adds an inbound Windows Firewall block rule for the attacker's address. |
step_1 | Shell (PowerShell) | After step_0, whatever its result | Disables the account that signed in. |
step_2 | Open incident | When step_1 succeeds | Opens an incident linked to the alert. |
step_3 | Send notification (ERROR) | When step_1 fails | Asks the team to disable the account by hand. |
- name: Windows account takeover response
description: Block the source, disable the account, open an incident.
conditions:
- operator: IS
field: name
value: 'Windows: Multiple Logon Failure Followed by Logon Success'
roots:
- step_0
nodes:
step_0:
kind: executor
executor: shell
platform: windows
shell: powershell
command: New-NetFirewallRule -DisplayName ('UTMStack block ' + $(alert.adversary.ip)) -Direction Inbound -RemoteAddress $(alert.adversary.ip) -Action Block
onSuccess:
- step_1
onError:
- step_1
step_1:
kind: executor
executor: shell
platform: windows
shell: powershell
command: net user $(alert.target.user) /active:no
onSuccess:
- step_2
onError:
- step_3
step_2:
kind: executor
executor: incident
params:
name: Account takeover $(alert.id)
description: Address $(alert.adversary.ip) blocked and account $(alert.target.user) disabled on $(alert.dataSource).
step_3:
kind: executor
executor: notify
params:
type: ERROR
message: Could not disable $(alert.target.user) on $(alert.dataSource) - act by handWhat happens. The firewall rule is named after the address, for example UTMStack block 203.0.113.45. The account is disabled whether or not the firewall rule worked, because step_1 is on both wires of step_0. The incident carries the alert's severity and links the alert.
Notes.
Incident names must be unique in a tenant, so the name includes
$(alert.id). In UTMStack v12.0.2 the inserted values appear in single quotes in the incident name and description.net user … /active:nodisables a local account, or a domain account when it runs on a domain controller.No placeholder is put inside quotes. UTMStack quotes each value for PowerShell.
Use case 3: Isolate a Windows machine on a ransomware sign, except critical servers
Scenario. Deleting shadow copies or backups is a common step just before ransomware encrypts files. You want to cut such a machine off the network at once, but never your domain controllers or database servers: for those, the team decides.
Trigger. Alert name is Windows: Ransomware Recovery Inhibition (Shadow Copy / Backup Deletion) (shipped rule).
Steps.
| Step | Type | Runs | What it does |
|---|---|---|---|
step_0 | Conditional | First | Checks that the machine is not in your list of critical servers. |
step_1 | Shell (PowerShell) | When the machine is not critical | Blocks all inbound and outbound traffic with Windows Firewall rules. |
step_2 | Send notification (ERROR) | When the machine is critical | Asks the team to act by hand. |
- name: Ransomware precursor - isolate workstation
description: Isolate the machine unless it is a critical server.
conditions:
- operator: IS
field: name
value: 'Windows: Ransomware Recovery Inhibition (Shadow Copy / Backup Deletion)'
roots:
- step_0
nodes:
step_0:
kind: executor
executor: conditional
params:
conditions:
- field: alert.dataSource
operator: IS_NOT_ONE_OF
value:
- DC01
- SQL01
onSuccess:
- step_1
onError:
- step_2
step_1:
kind: executor
executor: shell
platform: windows
shell: powershell
command: New-NetFirewallRule -DisplayName UTMStack-Isolate-Out -Direction Outbound -Action Block -RemoteAddress Any; New-NetFirewallRule -DisplayName UTMStack-Isolate-In -Direction Inbound -Action Block -RemoteAddress Any
step_2:
kind: executor
executor: notify
params:
type: ERROR
message: Ransomware sign on critical server $(alert.dataSource) - not isolated, act nowWhat happens. For a workstation, the conditional step matches, the green wire runs, and the machine loses all network access, including its connection to UTMStack. For DC01 or SQL01, the conditional step shows as Failed in Execution history (that only means the red branch was taken) and the team gets an error notification.
After step_1 runs, UTMStack can no longer reach the machine. Someone must remove the two UTMStack-Isolate firewall rules on the machine itself (at the console or through out-of-band management).
Notes.
Replace
DC01andSQL01with the host names exactly as they appear in the alert's Data source. Comparisons are case-sensitive.Keep the list in the Code tab as shown. If you type the values of a Conditional step's "is not one of" in the Visual tab, UTMStack v12.0.2 saves them as one piece of text and the condition always matches (see Step Types).
The isolation command is the palette's "Isolate host (block all traffic)" step for PowerShell.
Use case 4: Block an attacker on a Linux gateway from a network sensor alert
Scenario. Your Suricata sensor (an open-source network intrusion detection system) reports SSH brute force or tunnelling from an outside address. The address should be blocked on your Linux gateway, which runs a UTMStack agent, not on the sensor that raised the alert.
Trigger. Alert name is SSH Brute Force and Tunneling Detection (shipped Suricata rule).
Steps.
| Step | Type | Runs | What it does |
|---|---|---|---|
step_0 | Shell (bash), Run on > one host: your gateway | First | Drops forwarded traffic from the attacker's address. |
step_1 | Send notification (ERROR) | When step_0 fails | Asks the team to block by hand. |
- name: Block SSH attacker on the gateway
description: Block the source address of Suricata SSH alerts on the Linux gateway.
conditions:
- operator: IS
field: name
value: SSH Brute Force and Tunneling Detection
roots:
- step_0
nodes:
step_0:
kind: executor
executor: shell
platform: linux
shell: bash
agent: gw-01
command: iptables -I FORWARD -s $(alert.adversary.ip) -j DROP
onError:
- step_1
step_1:
kind: executor
executor: notify
params:
type: ERROR
message: Could not block $(alert.adversary.ip) on the gateway - act by handWhat happens. The command always runs on gw-01 (replace it with your gateway's host name, or pick the host in Run on > one host), whichever sensor raised the alert.
Notes.
If the gateway has no agent, use a jump host with an agent and run the command over SSH, with the key path and the gateway address in secret variables. See Running Steps on Another Host.
Check that the sensor fills in the adversary address for this rule before you switch the flow on.
Use case 5: Post an alert to a chat channel
Scenario. Your team works in a chat tool that accepts messages at an incoming-webhook address, for example Slack, which takes a JSON body with a text field. You want a message in the channel each time a ransomware sign is seen.
Trigger. Alert name is Windows: Ransomware Recovery Inhibition (Shadow Copy / Backup Deletion).
- name: Ransomware sign - post to chat
description: Post the alert to the security channel.
conditions:
- operator: IS
field: name
value: 'Windows: Ransomware Recovery Inhibition (Shadow Copy / Backup Deletion)'
roots:
- step_0
nodes:
step_0:
kind: executor
executor: http
params:
method: POST
url: https://hooks.example.com/your-webhook-address
headers:
Content-Type: application/json
body:
text: UTMStack alert $(alert.name) on $(alert.dataSource), severity $(alert.severity)What happens. The UTMStack server posts the message. The channel shows, for example, UTMStack alert 'Windows: Ransomware Recovery Inhibition (Shadow Copy / Backup Deletion)' on 'WKS-042', severity 'high'.
Notes.
The step's
kindisexecutor, because chat webhooks usually answer with plain text (such asok), which an enrichment step would reject.The request leaves from the UTMStack server, which must be able to reach the webhook address.
The webhook address works like a password. Anyone who can view flows can read it, and Execution history shows it in full.
For Microsoft Teams, ticketing tools and firewalls, see Worked Examples: Responding Through Other Systems.
In UTMStack v12.0.2 each inserted value arrives in single quotes, and a value containing an apostrophe or a double quote stops the step. Shipped rule names contain neither.
Use case 6: Let the AI decide, with a safe default
Scenario. For a noisy detection you want the SOC-AI assistant to look at the attacker's address first, and block it only when the assistant judges it malicious. Otherwise, a person should look.
Requirements. An AI provider must be set up (see AI Providers).
Trigger. Alert name is System Linux: Possible Brute Force Attack.
- name: AI-checked SSH block
description: Ask SOC-AI about the source address, block only when it says so.
conditions:
- operator: IS
field: name
value: 'System Linux: Possible Brute Force Attack'
roots:
- step_0
nodes:
step_0:
kind: enrichment
executor: llm_enrich
params:
prompt: Review the activity of address $(alert.adversary.ip) in the last 24 hours of logs. Set result to the single lowercase word block if it is clearly malicious, otherwise to the single lowercase word review.
onSuccess:
- step_1
onError:
- step_3
step_1:
kind: executor
executor: conditional
params:
conditions:
- field: step_0.result
operator: IS
value: block
onSuccess:
- step_2
onError:
- step_3
step_2:
kind: executor
executor: shell
platform: linux
shell: bash
command: iptables -I INPUT -s $(alert.adversary.ip) -j DROP
step_3:
kind: executor
executor: notify
params:
type: WARNING
message: SSH brute force from $(alert.adversary.ip) on $(alert.dataSource) needs reviewWhat happens. The assistant may search the logs before it answers. The conditional step reads its answer directly (step_0.result), so no quoting is involved. Only the exact answer block blocks the address; any other answer, or an AI error, sends a notification for a person to review.
Notes.
AI answers can vary. Keep the safe action (notify) on every branch that is not an exact match.
step_3has two incoming wires, but they can never both fire in one run (step_1runs only whenstep_0succeeds), so it runs at most once.An AI step can take minutes. A step still running after 5 minutes is stopped and marked Failed, so the error wire to
step_3sends the notification.Do not insert the AI answer into notifications, emails or HTTP calls in UTMStack v12.0.2: an apostrophe in the answer stops those steps (see Placeholders).
Combine the patterns
You can chain these patterns in one flow: block on the machine, block on the gateway, disable the account, open an incident and post to chat. Keep each step on a single incoming wire, because a step with several incoming wires runs once per wire (see Building a Flow).
Related pages
Building a Flow
Step Types
Placeholders
Step Outputs and Variables
Running Steps on Another Host
Shipped Flows
Worked Examples: Responding Through Other Systems
Troubleshooting Flows