Skip to content

Endpoint Requirements

This page lists the items that an endpoint must have to run the SenseOn Universal Sensor:

  • Supported operating systems
  • Hardware resources
  • Network connectivity
  • Special considerations for cloned hosts and for environments that use a proxy

Use this page for these tasks:

  • To make sure that a target estate is compatible before you deploy SenseOn.
  • To calculate the correct size of the hosts.
  • To give information to network and security teams who must add traffic to an allowlist or approve the agent.

💡 Scope of this page: These requirements apply to hosts where you plan to deploy the SenseOn Universal Sensor agent. Some hosts cannot run the agent (network appliances, IoT devices, OT equipment, third-party SaaS systems, or hosts that run unsupported operating systems). For these hosts, you can use Log Ingestion or deploy a Network Probe to bring telemetry into SenseOn. The three collection methods are complementary. You can use them together in one tenant.

💡 Do you want to know how it works? How the Universal Sensor Works gives an overview of the architecture. It tells you what the agent does on a host, how telemetry flows to the platform, and which controls protect that data.


Operating Systems

Each entry in the list that follows gives the architectures that SenseOn supports on that operating system.

  • Windows 8.1 and later desktop versions: 64 bit (x86_64) and ARM
  • Windows Server 2012 R2 and later server versions: 64 bit (x86_64) and ARM
  • Ubuntu Linux 18.04+: 64 bit (x86_64) only, no ARM
  • CentOS/RedHat 7+: 64 bit (x86_64) only, no ARM
  • macOS Catalina (10.15) or higher: Intel and Apple silicon

Supported Linux Distributions

SenseOn supplies packages for the distributions in the table that follows. On Linux, the architecture is x86_64 only (no aarch64 / ARM).

Distribution Officially supported versions Package format
Ubuntu 18.04 LTS, 20.04 LTS, 22.04 LTS, 24.04 LTS .deb
Debian Releases that correspond to the supported Ubuntu kernels .deb
Red Hat Enterprise Linux 7, 8, 9 .rpm
CentOS 7, 8 (Stream) .rpm
Rocky Linux 8, 9 .rpm
AlmaLinux 8, 9 .rpm
Amazon Linux 2 .rpm

Linux Kernel Requirements

On Linux, the Universal Sensor uses eBPF for telemetry in the kernel. The minimum kernel features are:

  • Linux kernel 5.2 or later for full functionality. This is the minimum version that supports BPF CO-RE (Compile Once, Run Everywhere). The sensor uses BPF CO-RE for eBPF programs that are portable across distributions.
  • Older kernels (4.x): the sensor installs and collects process telemetry. But the network telemetry that the sensor collects through eBPF kprobes is not available.

In practice, the results are:

  • Ubuntu 20.04 LTS (kernel 5.4) and newer give full functionality.
  • Ubuntu 18.04 LTS (kernel 4.15 by default) gives process telemetry only. The exception is a host that you upgraded to a Hardware Enablement (HWE) kernel of 5.2 or later.
  • RHEL / CentOS 8+ have kernel 4.18 in mainline but include backported eBPF support. RHEL / CentOS 9 has kernel 5.14.

🐧 Non-supported systems: It is possible that other Linux distributions, or kernels older than 5.2, are functional, but SenseOn does not officially support them. Network telemetry is available only with the kernel features that this section lists. If those features are not available, the sensor continues to collect process telemetry and the host continues to show in Assets > Devices.

Operating System Feature Differences

Where possible, SenseOn tries to give the same features on all platforms. But technical constraints, and the complexity of feature development for different environments, can prevent releases at the same time. The table that follows gives the state of the features on each platform.

Feature Windows macOS Linux
Network Detection and Response (NDR) Yes Yes Yes
Endpoint Detection and Response (EDR) Yes Yes Yes
Endpoint Protection (EPP) Yes Yes No
Active Response Yes Yes (Apple silicon only) Yes
Reflex Yes No No

CPU/RAM

SenseOn recommends a minimum of 2GB of RAM for systems that run SenseOn. The table that follows also gives the minimum requirements:

Operating System Recommended RAM Minimum RAM
Windows 4GB 512mb (Server Core), 2GB (Desktop)
macOS 4GB 2GB
Ubuntu 24+ 2GB 1GB
Debian 2GB 256mb

Impact on CPU/RAM

The performance of the endpoint agent changes with the modules that the agent loads.

  • On Windows with NDR and EDR features enabled but Endpoint Protection disabled:

    • Average CPU: <0.6% (depending on device type and hardware)
    • Average memory use: ~ 90 MB
  • On Windows with NDR, EDR, and Endpoint Protection features all enabled:

    • Average CPU: <1% (depending on device type and hardware)
    • Average memory use: ~ 300 MB

High CPU utilisation: On servers that are unusually busy and have EPP enabled, CPU utilisation can be as much as 20%. Examples are file servers and servers with a large number of reads/writes. The CPU utilisation changes with the workload.

💡 How to turn features on or off: To turn product features on or off, create a segment. Then change the configuration that applies to that segment. SenseOn must first enable licensed modules such as Endpoint Protection and Media Control for your organisation.

Network Requirements

Bandwidth

The sensor uses 28 kilobits per second (~300 MiB in 24 hours). This value changes with device activity and workload.

💡 Data Latency: The platform ingests data in real-time. A delay of not more than tens of seconds can occur while the platform analyses the data, before the user interface shows the data. This delay has no effect on the automatic actions of the EPP, which occur immediately.

Connectivity Requirements

Endpoint sensors that use the cloud-hosted analysis platform communicate with the domains that follow.

💡 Specific domains: If necessary, SenseOn can give you specific domain names for your tenant. To get them, contact support.

Domain Name Purpose
<your-tenant>.senseon.io Hosts the install script that the install command gets. This domain is necessary to install and update the sensor.
*.snson.net Endpoint agent telemetry and active response sessions.
avmirror.snson.net Antivirus updates that EPP uses. The wildcard in the row before this one includes this domain.
*.s3.<your-region>.amazonaws.com Uploads/downloads during active response and crash dumps. Each active response session uses a unique bucket. If you do not use active response, this domain is not necessary.
seev4.s3.<your-region>.amazonaws.com Endpoint agent updates. The wildcard in the row before this one includes this domain. This domain is not necessary if all agents on the estate are v7.0 or later and the bootstrapper installed them. In that condition, the bootstrapper supplies the agent updates.

⚠ Allow the random callback subdomain. Each SenseOn deployment calls back on a random subdomain under snson.net (the subdomain matches *.snson.net). If your egress policy cannot allow the full wildcard and must have the exact subdomain, contact support to get the unique value for your tenant.

⚠ Use the regional S3 endpoints. <your-region> is the AWS region that hosts your tenant, for example eu-west-1 or eu-west-2. For a tenant in eu-west-1, the update domain is seev4.s3.eu-west-1.amazonaws.com. The form without the region, seev4.s3.amazonaws.com, does not match, because the sensor connects to the regional endpoint. If you do not know which region applies to your tenant, contact support.

💡 IP addresses: The IP address for endpoint agent callbacks stays static unless the region of the tenant changes. The IP addresses at *.s3.<your-region>.amazonaws.com are dynamic. You can filter them only with domain names.

TLS Interception

Endpoint sensors use Mutual TLS 1.2 or later to communicate with the analysis platform. TLS interception (SSL inspection) breaks the installation and the connectivity after installation. If you use TLS interception, add a decryption bypass on the inspection appliance for the domains in the table that follows. The label of a decryption bypass is frequently "do not decrypt" or "SSL bypass".

Domain Name Purpose
*.snson.net Endpoint agent telemetry and active response sessions.
*.senseon.io Install script downloads.
seev4.s3.<your-region>.amazonaws.com Update package downloads. Necessary only for agents on v6.x and earlier.
*.s3.<your-region>.amazonaws.com Crash diagnostic uploads and active response file transfers, on every agent version.

The two S3 entries are for different traffic. You cannot use one in place of the other:

  • Update packages. Agents on v6.x and earlier download these directly from the seev4 bucket. If a bypass includes only *.snson.net and *.senseon.io, those agents cannot update, but telemetry continues to work normally. This is the most common symptom of this problem. Agents on v7.0 and later download update packages from the appliance. Thus the seev4 entry is not necessary for them.
  • Crash diagnostics and active response. On every agent version, v7.0 and later included, these go directly to S3. Each active response session uses its own bucket. You cannot know the bucket name of a session before the session starts. Thus only the *.s3.<your-region>.amazonaws.com wildcard includes these buckets. This entry is also necessary for an estate that is fully on v7.0. Without it, these functions do not work, although updates are successful.

If your policy supports wildcards, *.s3.<your-region>.amazonaws.com alone is sufficient for the two cases.

These are some of the signs that TLS interception blocks SenseOn:

  • Installations stop with Windows Installer error 1603.
  • Endpoints install but never connect.
  • A part of the estate does not update to the latest version, but the other part does.

The last sign shows that interception blocks the update on the affected hosts.

⚠ Do not rely on trust of your inspection CA. If you add the certificate authority of the inspection appliance to the operating system trust store, the update failures continue. The sensor does a check of the update download against its own bundled certificate list, not against the OS trust store. Thus a decryption bypass is necessary.

⚠ Apply the bypass at each location where inspection occurs. You configure the bypass on your own proxy. SenseOn does not configure it. Thus the bypass must be present at each location or policy that enforces TLS inspection. Some endpoints connect to the platform through an inspected path without the bypass. On those endpoints, the installation does not work, or the endpoint loses connectivity after the installation.

Antivirus and EDR Exclusions

Third-party antivirus and EDR products can interfere with SenseOn. They can block the installer and the bootstrapper before the installation is complete, or they can quarantine sensor components at runtime. If you deploy SenseOn together with a different security product, add exclusions in that product for the SenseOn install locations and processes. If possible, apply the exclusions before you install SenseOn.

An exclusion for the install directories in the tables that follow includes each SenseOn executable in those directories. If a product also applies process or behavioural exclusions, also exclude the process names in the tables.

Windows

Type Value
Directory C:\Program Files\senseon-see\
Directory C:\ProgramData\senseon-see\
Directory C:\Program Files\senseon-bootstrapper\
Directory C:\ProgramData\senseon-bootstrapper\
Process senseon-seed.exe
Process senseon-bootstrapper.exe, senseon-broker.exe, senseon-logger.exe

Linux

Type Value
Directory /etc/senseon-see/, /var/senseon-see/, /var/log/senseon-see/, /usr/share/senseon-see/
Directory /var/senseon-bootstrapper/, /var/log/senseon-bootstrapper/
Binary /usr/bin/senseon-seed, /usr/bin/senseon-seei, /usr/bin/senseon-epns
Binary /usr/bin/senseon-bootstrapper, /usr/bin/senseon-broker, /usr/bin/senseon-logger, /usr/bin/senseon-token

macOS

Type Value
Directory /opt/senseon-see/, /var/senseon-see/
Directory /opt/senseon-bootstrapper/
launchd label io.senseon.endpoint-sensor, io.senseon.notification
launchd label io.senseon.bootstrapper, io.senseon.broker, io.senseon.logger

💡 Microsoft Defender for Endpoint: Add the directories in the tables as path exclusions and the processes as process exclusions. SenseOn recommends that you do this before problems occur on estates that run Defender together with SenseOn. If you do not, it is possible that Defender interferes with the agent or makes the installation slow.

💡 Windows Server: Exclusions do not make Windows Security show SenseOn as an enabled antivirus product on a server. Server editions do not have that reporting mechanism. The absence of that mechanism has no relation to the protection that SenseOn gives the host. Where SenseOn is the antivirus and EDR of a server, Windows Defender Antivirus stays active together with SenseOn until you act. Windows Defender on Windows Server gives the supported procedure to stop Windows Defender Antivirus. That procedure applies to servers that are onboarded to Microsoft Defender for Endpoint and to servers that are not.

⚠ Apply the exclusions before deployment where possible. The installer package downloads to a temporary location and runs from there. If your antivirus quarantines the downloaded installer before it runs, the installation does not start. To prevent blocked installations, add these exclusions before a deployment.

Golden Images

An identity connects each SenseOn endpoint agent to the host on which you installed it. The installation makes this identity. When you clone a host, each cloned host must have its own SenseOn identity. This applies to Citrix MCS / PVS, VMware Horizon, AWS AMI, Azure image, and all other golden-image workflows. If the clones do not have their own identities, they collide in Assets > Devices, and it is possible that hosts go missing.

The simplest reliable procedure is to keep SenseOn off the golden image and run the install command on the first boot of each cloned instance.

  1. Do not install SenseOn on the golden image / template. Build the image with your usual application stack, but do not include the SenseOn agent.
  2. Configure your provisioning system to run the SenseOn install command on first boot. The install command is available in Settings > Universal Sensor. Each clone runs the installer one time and registers as a different device.
  3. Seal or capture the image with your usual process.

With this pattern, deletion of individual files or an identity reset on the template is not necessary. Each clone gets a new installation of the agent, thus a clone cannot inherit an identity from the image.

Running the install command on first boot

The install command and the installer key for your tenant are available from Settings > Universal Sensor. In the examples that follow, replace <your-tenant> and <your-installer-key> with your own values.

Use the first-boot mechanism that your provisioning platform already supports. Examples:

Windows (Citrix MCS / PVS, VMware Horizon, generic)

Configure a startup script, scheduled task, or Group Policy startup item. It must run the SenseOn install command in an elevated PowerShell session:

$env:SENSEON_INSTALLER_KEY="<your-installer-key>"; `
  iwr -useb https://<your-tenant>.senseon.io/install.ps1 | iex

For machines that Citrix provisions, put the script in the startup-script hooks of the Citrix machine catalogue. Alternatively, start it through your standard Group Policy startup-script configuration. Deployment with Group Policy (GPO) gives a step-by-step guide to the Group Policy procedure. The guide includes an installed-check, which makes the script safe to run at each boot.

Windows (sysprep)

If you use sysprep, add the install command to the FirstLogonCommands (or SynchronousCommand for unattended setup) section of your unattend.xml:

<FirstLogonCommands>
  <SynchronousCommand wcm:action="add">
    <Order>1</Order>
    <CommandLine>powershell.exe -ExecutionPolicy Bypass -Command "$env:SENSEON_INSTALLER_KEY='&lt;your-installer-key&gt;'; iwr -useb https://&lt;your-tenant&gt;.senseon.io/install.ps1 | iex"</CommandLine>
    <Description>Install SenseOn agent on first boot</Description>
  </SynchronousCommand>
</FirstLogonCommands>

AWS EC2 (AMI-based instances)

Add the install command to the instance user data. Then the command runs on first boot:

#!/bin/bash
curl -fsSL https://<your-tenant>.senseon.io/install.sh \
  | SENSEON_INSTALLER_KEY="<your-installer-key>" bash

Refer to Deploying the SenseOn Universal Sensor on AWS EC2 for full details, which include Auto Scaling launch templates.

Azure VM

Use a Custom Script Extension or cloud-init user data on the VM to run the same curl ... | bash command as in the AWS EC2 example.

Linux (cloud-init / systemd)

Add a one-shot systemd service. The service runs the installation at first boot and then disables itself:

[Unit]
Description=SenseOn first-boot install
ConditionPathExists=!/etc/senseon-see/.first-boot-done

[Service]
Type=oneshot
Environment="SENSEON_INSTALLER_KEY=<your-installer-key>"
ExecStart=/bin/bash -c "curl -fsSL https://<your-tenant>.senseon.io/install.sh | bash"
ExecStartPost=/usr/bin/touch /etc/senseon-see/.first-boot-done

[Install]
WantedBy=multi-user.target

macOS (Jamf Pro / Kandji / generic MDM)

Use the policy mechanism of your MDM to run the macOS install command on each newly enrolled device. Refer to Deployment with Jamf Pro and Deployment with Kandji for the instructions for each tool.

Verifying a Cloned Host Has Registered

After you provision a cloned host:

  1. Log in to SenseOn.
  2. Go to Assets > Devices.
  3. Make sure that the new host shows in the list with its expected hostname and a recent Last Seen timestamp.

This check is most important in VDI environments such as Citrix XenApp / XenDesktop, where you provision many short-lived hosts from the same image. If a host does not show in 30 minutes or less after it comes online, contact support. Send the affected hostname(s) and a short description of your provisioning flow to support@senseon.io.

If you must pre-install on the golden image

If the endpoints have no direct internet but can connect to SenseOn through a proxy, this procedure is not necessary. Install through the proxy with the -Proxy option of the install script. Installing Through or Around a Proxy describes this option.

In some environments, hosts cannot connect to the install-script URL of SenseOn on first boot. Examples are air-gapped subnets with no proxy path, or strict egress policies that start before your first-boot script runs. In these environments, you can pre-install on the golden image, but you must be more careful. Contact support@senseon.io. SenseOn will give you the steps to reset the identity of the agent at clone time.

Proxy Configuration

The endpoint sensor supports a TLS pass-through HTTPS proxy. To enable the proxy, deploy a JSON file in the specified path. The sections that follow give the configuration details.

ℹ Note: This file sets the proxy that the running sensor uses for its continuous connection to the platform. On Universal Sensor 7.0.50 and later, the --proxy option of the install script writes this file for you, and --direct removes it. Create or edit the file manually to change the proxy after installation. Also do this on hosts that you installed without --proxy. Also do this on earlier versions, where --proxy applies only to the installation and to the first contact of the bootstrapper. Refer to Installing Through or Around a Proxy.

⚠ Use the path that applies to this host. There are two possible paths, and the correct one depends on whether this host has a bootstrapper. The instructions that follow use the canonical path. The bootstrapper, and a sensor that the bootstrapper manages, read this path. An older, sensor-only host with no bootstrapper does not read that path. Such a host must use the legacy path: C:\ProgramData\senseon-see\proxy.json (Windows) or /var/senseon-see/proxy.json (Linux and macOS). If you do not know which path applies to this host, refer to Troubleshooting: Is the bootstrapper installed?.

Windows

  1. Stop the endpoint sensor service.
    1. To open Start, click the Windows symbol in the bottom left corner or press the Windows Key.
    2. Search for Services (or services.msc).
    3. Click the top result to open the console.
    4. Double-click the service with the name senseon-seed.
    5. Click the Stop button.
    6. Look at the list of services to make sure that the service is stopped.
  2. Go to C:\ProgramData\senseon-bootstrapper\. To see C:\ProgramData and other hidden folders, go to File Explorer → View and click Show hidden files and folders.
  3. Open the file C:\ProgramData\senseon-bootstrapper\proxy.json. If the file does not exist, create it.
  4. Write the line that follows into the file, with the applicable values in the host and port fields:
    {"proxy_host": "127.0.0.1", "proxy_port": 8118}
    
  5. Save the file.
  6. Start the endpoint sensor service.
    1. To open Start, click the Windows symbol in the bottom left corner or press the Windows Key.
    2. Search for Services (or services.msc).
    3. Click the top result to open the console.
    4. Double-click the service with the name senseon-seed.
    5. Click the Start button.
    6. Look at the list of services to make sure that the service runs.

Linux

  1. Run the command systemctl stop senseon-seed to stop the endpoint sensor service.
  2. Go to /var/senseon-bootstrapper.
  3. Open the file /var/senseon-bootstrapper/proxy.json. If the file does not exist, create it.
  4. Write the line that follows into the file, with the applicable values in the host and port fields:
    {"proxy_host": "127.0.0.1", "proxy_port": 8118}
    
  5. Save the file.
  6. Run the command systemctl start senseon-seed to start the endpoint sensor service.

macOS

  1. Run the command sudo seectl stop to stop the endpoint sensor.
  2. Go to /var/senseon-bootstrapper.
  3. Open the file /var/senseon-bootstrapper/proxy.json. If the file does not exist, create it.
  4. Write the line that follows into the file, with the applicable values in the host and port fields:
    {"proxy_host": "127.0.0.1", "proxy_port": 8118}
    
  5. Save the file.
  6. Run the command sudo seectl start to start the endpoint sensor.