> For the complete documentation index, see [llms.txt](https://docs.vectra.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.vectra.ai/deployment/getting-started/ndr-network-identity-architecture.md).

# NDR / Network identity architecture

High-level architecture overview for Vectra NDR and network identity.

This page gives a high-level architecture overview of a Vectra NDR (Detect) deployment, including network identity coverage. Use it alongside the [Respond UX deployment guide](https://docs.vectra.ai/deployment/getting-started/respond-ux-deployment-guide.md) or the [Quadrant UX deployment guide](https://docs.vectra.ai/deployment/getting-started/quadrant-ux-deployment.md) when planning where to place appliances and what traffic to capture.

## Deployment Types: Respond UX and Quadrant UX

Vectra offers two analyst user experiences. See [Analyst UX options (RUX vs QUX)](https://docs.vectra.ai/deployment/getting-started/analyst-ux-options-rux-vs-qux.md) for a full comparison.

* **Respond UX (RUX)** – The Vectra UI is served from Vectra's cloud and is the central point of web-based management for RUX deployments. When used with network Sensors, it communicates with Brain appliance(s) deployed in the customer premises.
* **Quadrant UX (QUX)** – The Vectra UI is served from Brain appliance(s) and is the central point of web-based management for QUX deployments.

## Key Definitions

* **Vectra Cloud** – The portions of the Vectra AI Platform that reside in Vectra's cloud.
* **Customer premises** – Private/shared data centers, public cloud, and campus/office environments where Brain or Sensor appliances are installed to capture network traffic (including network identity).
* **Appliance** – A physical or virtual (including public IaaS cloud) Brain, Sensor, or Stream appliance.
* **Vectra AI Platform** – Vectra's cloud platform, delivered as SaaS with the RUX UI.
* Deployments with network Sensors include at least one Brain appliance and one or more Sensors. For smaller deployments, this can be a single mixed-mode appliance.

#### Brain Appliance

* Can be a physical or virtual appliance.
* Pairs with Sensors (network data sources) and processes and deduplicates the metadata received from Sensors, optionally forwarding it when licensed for [Stream](https://docs.vectra.ai/deployment/stream.md) or [Recall](https://docs.vectra.ai/deployment/recall-qux-only.md) (Recall is for QUX deployments only).
* Serves as the communications broker between Vectra's cloud and local integration points for RUX deployments.

#### Sensor Appliance

* vSensor is a virtual Sensor (for hypervisors or IaaS cloud); Sensor is a physical Sensor.
* Must be paired to a Brain.
* Captures and deduplicates raw network traffic, then forwards metadata to the Brain for processing.
* Houses a rolling capture buffer to enable PCAP retrieval when requested from the Brain.
* Optionally runs [Vectra Match](https://docs.vectra.ai/deployment/match.md) and [Suspect Protocol Anomaly Detections](https://docs.vectra.ai/operations/general/suspect-protocol-activity-detections-feature-overview.md).

#### Mixed-Mode Appliance

Can perform both Brain and Sensor functions when used in smaller deployments.

#### Network Identity

Vectra covers IDR (Identity Detection and Response) use cases through analysis of identity as observed in the network through protocols such as Kerberos, DCE/RPC, LDAP, NTLM, and others.

## Getting Started with a Deployment

{% hint style="info" %}
Knowledge Base (KB) articles, support cases, and product downloads are hosted on the [Vectra Support Portal](https://support.vectra.ai/s/). Formal product documentation, including deployment and configuration guides, lives here on this documentation site — use the navigation on the left side of this page to browse by topic.
{% endhint %}

* **RUX** – Receive your welcome email for a RUX deployment, log in and configure user accounts/SSO, deploy your Brain appliance, perform configuration/integration, deploy Sensors, and direct traffic to the Sensors.
* **QUX** – This deployment is largely the same, except there is no RUX UI, and your UI is served from the Brain.

Vectra offers a full complement of services including technical support, professional services/training, managed services, and offensive security services. See the [Vectra AI Services datasheet](https://www.vectra.ai/resources/vectra-ai-services) and speak with your Vectra account team for details and pricing.

After deployment, enable backups and integrations, enable other features/products, configure notifications/reporting, and build groups and triage rules to suppress unwanted detections for authorized behaviors. AI-Triage also learns the environment and automatically suppresses detections as part of [AI-driven Prioritization](https://docs.vectra.ai/reference/ai-driven-priortization-faq.md).

## General Deployment Architecture

The diagram below represents a high-level Respond UX deployment of Vectra NDR (including network identity).

<figure><img src="https://4227135129-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FHJ1ltuWFvsArFWtevnRn%2Fuploads%2Fgit-blob-46b4e39f7505adb50c2dcc75572093b4f14bb2a6%2Fndr-network-identity-architecture-1.png?alt=media" alt="Diagram of a Respond UX deployment showing Sensors in IaaS public cloud, datacenter, and plant/OT/ICS/IoT/campus environments sending packet metadata to a Brain in the customer&#x27;s Security Operations environment, which exchanges data with the Vectra Data &#x26; Analytics Cloud and forwards data to Respond UX, SOAR/SIEM, and 24x7 MDR/training via API."><figcaption></figcaption></figure>

* A Brain appliance is installed in the customer premises, and the RUX UI is served from the Vectra cloud.
* Sensors are paired to the Brain and capture network traffic in the data center, campus, and IaaS public clouds. Sensors can also run Vectra Match and Suspect Protocol Anomaly Detections.
* The Brain appliance processes the metadata locally to create detections, then sends metadata and detections to the Vectra cloud for presentation in the Respond UX.
* For QUX deployments, the UI is local, and metadata can be sent to [Vectra Recall](https://docs.vectra.ai/deployment/recall-qux-only.md).
* [Vectra Stream](https://docs.vectra.ai/deployment/stream.md) sends network metadata to the customer's data lake for both RUX and QUX deployments.
* Logs are analyzed by the Brain to link remote workers with traffic seen by app connectors in the data center.
* SOAR/SIEM integration uses API connections to communicate with the Vectra cloud.

## Placement of Brain and Sensors

* It is generally recommended to deploy Brain and Sensor devices in locations not visible from the public internet. Private connectivity or a VPN tunnel terminated outside of the Vectra appliances is preferred.
* Small remote branch locations do not all typically need a Sensor — flows are still seen in central locations.
* Vectra NDR detections are focused on in-to-in and in-to-out communications. It is preferred to not capture outside of your edge firewalls in the DMZ, where a significant amount of traffic will be out-to-in. Out-to-out traffic should not be captured. Vectra Match or some metadata use cases may necessitate out-to-in traffic capture.
* Vectra allows you to define your internal IP space. Exceptions can be configured for internal labs where you may wish to simulate C2 traffic, for example. Networks can be fully ignored by VLAN or CIDR block.

{% hint style="info" %}
**Proxy guidance** — see [Proxies](https://docs.vectra.ai/configuration/setup/proxies.md) for full details.

* Traffic should be captured on the south side of any proxy or NAT device to fully recognize the real sending host. In some situations, you may need to capture traffic on the north side of proxies.
* Flows from multiple hosts behind the proxy will be mapped to the north-side proxy IP and would result in spurious detections. Vectra automatically suppresses detections that would fire inaccurately because of this. The *Configuration → SETUP → Proxies* page in your Vectra UI lets you manage identified proxies.
  {% endhint %}

## Network Traffic Capture Guidance

### General Traffic Types to Capture

| Traffic type | Purpose                  | Examples                                       |
| ------------ | ------------------------ | ---------------------------------------------- |
| North/South  | C2, exfiltration, botnet | Server to internet, user to internet           |
| East/West    | Recon, lateral movement  | Server to server, user to server, user to user |

{% columns %}
{% column %}

### Important to Capture

* DCE/RPC
* DHCP
* DNS
* HTTP
* ICMP
* Kerberos
* LDAP
* NTLM
* RADIUS
* RDP
* SMB
* SMTP
* SSH
* SSL/TLS
* X.509
* Other session traffic
  {% endcolumn %}

{% column %}

### Should Be Excluded from Capture

* Core routing protocols
* High-bandwidth backup data
* High-performance computing (HPC) workloads high in bandwidth or well isolated
* Multiprotocol Label Switching (MPLS)
* Session Initiation Protocol (SIP)
* Storage network file systems (SMB is fine to capture)
* Video multicast
  {% endcolumn %}
  {% endcolumns %}

### What Helps Improve Vectra's HostID

See [Host ID best practices and functionality](https://docs.vectra.ai/reference/understanding-vectra-host-naming.md) for more information.

* DNS, reverse DNS, and multicast DNS (mDNS)
* Kerberos
* DHCP
* NetBIOS
* EDR integration, VMware integration, and SIEM event forwarding
* [Windows Event Log Ingestion (WELI)](https://docs.vectra.ai/configuration/coverage/network-identities-weli/windows-event-log-ingestion-weli.md)

{% columns %}
{% column %}

### Supported Encapsulations

* IEEE 802.1Q / IEEE 802.1ad (VLAN/QinQ)
* Generic Routing Encapsulation (GRE)
* Encapsulated Remote SPAN (ERSPAN)
* Virtual Extensible LAN (VXLAN)
* GENEVE
* IPsec Authentication Header (IPsec AH)
  {% endcolumn %}

{% column %}

### Typical Traffic Capture Sources

* SPAN/COPY/MIRROR ports
* Traditional network TAP devices
* Packet brokers
* Native cloud mirroring, such as [VPC Traffic Mirroring (AWS)](https://docs.aws.amazon.com/vpc/latest/mirroring/what-is-traffic-mirroring.html), [Packet Mirroring (GCP)](https://cloud.google.com/vpc/docs/packet-mirroring), and [VTAP (Azure)](https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-tap-overview)
  {% endcolumn %}
  {% endcolumns %}

## Encrypted Traffic

Vectra AI models work with encrypted traffic — decryption is not required for the vast majority of Vectra AI network traffic-related detections. For more information, see the whitepaper "The AI Behind Vectra Attack Signal Intelligence" (the relevant section begins on page 12) on the [Vectra AI resources site](https://www.vectra.ai/resources/the-ai-behind-vectra-ai).

Some customers may still have a need for decrypted data:

* Some users of [Vectra Recall](https://docs.vectra.ai/deployment/recall-qux-only.md), [Vectra Stream](https://docs.vectra.ai/deployment/stream.md), Instant Investigation, or Advanced Investigation may require decrypted traffic metadata. These use cases can vary but are not very common, and are supported.
* If using [Vectra Match](https://docs.vectra.ai/deployment/match.md), decryption can be useful and allow certain rules to fire that would otherwise not fire if the system processed only encrypted data. This requirement will vary by customer implementation.

{% hint style="info" %}
**Guidelines when decrypting traffic sent for analysis to Vectra Sensors**

* When you want both encrypted and decrypted traffic to be analyzed, this should be done in parallel pipelines. A second Sensor (or as many as are needed for your specific throughput or architecture requirements) should be used, paired to a different Brain appliance.
* Vectra supports virtual Sensors, Brains, and Stream appliances, and does not charge for their use outside of normal licensing metrics for your environment.
* If you do not use a parallel processing pipeline and send both decrypted and encrypted traffic to the same Sensors and paired Brain, Vectra's deduplication will see the same 5-tuple traffic in multiple data streams and will discard the duplicate data that arrives last. This is typically the decrypted data, since the decryption process adds overhead on the sending side.
  {% endhint %}

## Basic Communications and Pairing

<figure><img src="https://4227135129-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FHJ1ltuWFvsArFWtevnRn%2Fuploads%2Fgit-blob-6715e54006e77193c349109913e48c8c6adad4d3%2Fndr-network-identity-architecture-2.png?alt=media" alt="Diagram showing Respond UX connecting to the Vectra cloud over TCP/443, the Vectra cloud connecting to the Brain (connections initiated from the Brain), and vSensors and physical Sensors connecting to the Brain over TCP/22 and TCP/443."><figcaption></figcaption></figure>

* Users log in with their web browser to the Respond UX at the URL given in their welcome letter. Quadrant UX users log in to their Vectra UI (served from the Brain) at the IP or hostname of their Brain.
* Sensors must be able to reach the Brain over TCP/22 (SSH) and TCP/443 (HTTPS). In general, Sensors only communicate with the Brain appliance and DNS, with one exception: physical Sensors can retrieve the IP of the Brain they need to pair with from the Vectra Cloud at initial boot. With DHCP enabled, a physical Sensor can automatically begin the pairing process with no login required to its CLI — this requires that TCP/443 be open to `update2.vectranetworks.com` from the physical Sensor.
* The Brain appliance is deployed in the customer premises and communicates with the Vectra Cloud over a variety of channels. These connections are initiated from the Brain.
* Updates are served from the Vectra Cloud and installed automatically in an update window chosen by the customer. Sensor updates are served from the Brain appliance and are also installed automatically.
* Pairing Sensors or Stream appliances with a Brain is a simple process — see [Pairing appliances](https://docs.vectra.ai/deployment/appliance-operations/pairing-appliances.md), along with [NDR physical appliances](https://docs.vectra.ai/deployment/ndr-physical-appliances.md) or [NDR virtual / cloud appliances](https://docs.vectra.ai/deployment/ndr-virtual-cloud-appliances.md) for the deployment guide specific to your hardware, hypervisor, or cloud provider. A Sensor registration token, managed in your UI or CLI, allows IaaS cloud vSensors or any other Sensor to pair with a Brain even if it was not originally configured for use with a specific Brain.

{% hint style="info" %}
See [Firewall requirements](https://docs.vectra.ai/deployment/getting-started/firewall-requirements.md) for full details — there are additional requirements depending on the specifics of your deployment and integrations.

Air-gapped deployments of Vectra NDR are possible (QUX only). NDR detections are generated locally on the Brain. See [Offline updates (v8.9+)](https://docs.vectra.ai/operations/readme-1/offline-updates-v89.md) for details on how updates are managed in these situations.
{% endhint %}

## Appliance Sizing

{% hint style="info" %}
Vectra periodically releases new appliances to support new throughput requirements, hypervisors, or cloud providers. Always refer to [Appliance specifications](https://docs.vectra.ai/deployment/getting-started/appliance-specifications.md) and your Vectra account team for the latest guidance.
{% endhint %}

* Brain appliances are available in configurations that support up to 100 Gbps of performance and 500 paired Sensors.
* Sensor appliances are available in configurations that support up to 100 Gbps of performance.
* When running Match and/or Suspect Protocol Anomaly Detections in addition to Vectra NDR (Detect), Sensor appliances are available in configurations that support up to 40 Gbps of performance.
* Performance refers to the amount of network traffic observed that a Sensor can produce metadata for, or that a Brain can process metadata for, based on average throughput. Actual performance may vary depending on traffic composition. Work with your Vectra account team to determine the right mix of physical and virtual appliances required.

See [Appliance specifications](https://docs.vectra.ai/deployment/getting-started/appliance-specifications.md) for detailed per-model throughput and sizing information.

## Global View

Global View enables maximum scale for the largest global enterprises. The performance guidelines above are for individual appliances — Global View provides a Global Respond view of prioritized entities from child RUX instances in an anchor instance of RUX.

<figure><img src="https://4227135129-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FHJ1ltuWFvsArFWtevnRn%2Fuploads%2Fgit-blob-059d79bb47283fa721f0e8e95de25e75b7f43402%2Fndr-network-identity-architecture-3.png?alt=media" alt="Diagram showing an anchor RUX instance retrieving data from three child Vectra instances, each with network, identity, and public cloud detection sources, alongside a Global Respond screenshot showing prioritized entities filtered across Vectra instances."><figcaption></figcaption></figure>

* Available as a standard feature with any RUX deployment.
* Allows for overlapping IP space in child deployments, enabling maximum scale while still providing a single pane of glass.
* The anchor instance retrieves prioritized entities from child instances using an API client assigned the "Global Analyst" role.
* Communications are all encrypted and contained within the Vectra AI Platform. Data is not stored in the anchor instance — it is retrieved on demand from child instances.

See [Global View](https://docs.vectra.ai/deployment/getting-started/respond-ux-deployment/global-view.md) for setup details.

## Additional Guidance and Resources

* Detection data (including associated micro-PCAPs) is generally retained for 6 months. Metadata retention depends on your contracted retention options for Event Search (Investigate) (RUX) or Recall (QUX). When sending metadata via Stream, retention is based on the configuration of your downstream data lake.
* [Using Vectra packet capture (PCAP)](https://docs.vectra.ai/deployment/traffic-engineering-and-validation/using-vectra-packet-capture-pcap.md) lets you capture PCAPs for subsequent download and analysis with local tools.
* Guidance for optimizing your deployment when used with VPN clients is available in [Optimizing Vectra for use with VPN clients](https://docs.vectra.ai/configuration/coverage/remote-users/optimizing-vectra-for-use-with-vpn-clients.md). Zscaler-specific information is available in [Zscaler ZPA](https://docs.vectra.ai/configuration/coverage/remote-users/zscaler-zpa.md) and [Zscaler ZIA](https://docs.vectra.ai/configuration/coverage/remote-users/zscaler-zia.md) — see [*Integrations → Coverage*](https://docs.vectra.ai/integrations/coverage.md) for other remote-user and cloud coverage integrations.
* **Response options** (not a complete list — see [*Integrations → Response*](https://docs.vectra.ai/integrations/response.md) for all supported response and lockdown integrations):
  * [Host Lockdown (EDR)](https://docs.vectra.ai/configuration/response/lockdown/host-lockdown-edr.md)
  * [Active Directory Account Lockdown](https://docs.vectra.ai/configuration/response/lockdown/active-directory-account-lockdown.md) and [Entra ID (Azure AD) Account Lockdown (RUX)](https://docs.vectra.ai/configuration/response/lockdown/entra-id-azure-ad-account-lockdown-rux.md)
  * [Vectra Automated Response](https://github.com/vectranetworks/vectraautomatedresponse) — a GitHub repo with dozens of integrations
* **SOAR** — see [*Integrations → SOAR*](https://docs.vectra.ai/integrations/soar.md) for supported platforms, including Splunk SOAR, Palo Alto Networks XSOAR, ServiceNow SIR, and Google SecOps SOAR.
* **SIEM and data export** — see [*Integrations → SIEM*](https://docs.vectra.ai/integrations/siem.md) for supported platforms (Splunk, IBM QRadar, Microsoft Sentinel, Google Chronicle, and more) and [*Integrations → Notifications / Data Export*](https://docs.vectra.ai/integrations/notifications-data-export.md) for webhook, syslog, and Kafka options. Syslog/Kafka is not supported in RUX deployments but is available for QUX deployments; if you require Syslog for RUX, this can be done through a customer-deployed intermediary server — see [SIEM connector (syslog intermediary)](https://docs.vectra.ai/deployment/getting-started/respond-ux-deployment/siem-connector-syslog-intermediary.md), which pulls data via API and sends it to your SIEM.
* Vectra is a very open platform. API guides are available for [Respond UX deployments](https://docs.vectra.ai/configuration/access/api-rux.md) and [Quadrant UX deployments](https://docs.vectra.ai/configuration/access/api-qux.md).
* MITRE ATT\&CK and D3FEND details are available in [Vectra's coverage of MITRE ATT\&CK and D3FEND](https://docs.vectra.ai/reference/vectra-coverage-of-mitre-attandck-and-d3fend.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.vectra.ai/deployment/getting-started/ndr-network-identity-architecture.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
