Featured on Product HuntReview on Product Hunt
← All posts
Networking Security

East-West Traffic Monitoring

NovaSafe Team·

A network can have strong defenses at its internet boundary and still leave important activity unseen. Once an attacker or misconfigured service is inside, communication between internal systems ca…

Share

A network can have strong defenses at its internet boundary and still leave important activity unseen. Once an attacker or misconfigured service is inside, communication between internal systems can help them discover assets, steal credentials, move laterally, or reach sensitive data. This internal communication is known as east-west traffic.

East-west traffic monitoring helps security and operations teams understand those connections and identify activity that deserves investigation. It is not a single tool or deployment pattern: teams can use network sensors, flow records, endpoint telemetry, cloud-native logs, or a combination. Each approach offers a different balance of detail, coverage, cost, and operational effort.

This guide explains what east-west monitoring can reveal, compares common approaches, and offers criteria for designing coverage that fits your environment.

What is east-west traffic?

East-west traffic is communication between workloads or systems within an environment. Examples include:

  • An application server querying a database.
  • A virtual machine connecting to an internal identity service.
  • Containers communicating across nodes in a cluster.
  • A cloud workload calling an internal API in another subnet.
  • An administrator connecting to a server over a management network.

By contrast, north-south traffic generally describes communication entering or leaving an environment, such as a user request arriving from the internet or a workload connecting to an external service. The distinction is useful, but not absolute: network boundaries and routing designs vary, and communication between cloud environments, regions, or data centers may not fit neatly into either category.

Monitoring east-west traffic means collecting and analyzing evidence about these internal connections. Depending on the method, that evidence may include source and destination addresses, ports, protocols, connection timing, transferred bytes, DNS requests, application metadata, or—in some cases—packet contents.

Why monitor internal traffic?

Internal networks should not be treated as inherently trusted. An endpoint may be compromised, a service may be exposed by mistake, or a legitimate account may be abused. Monitoring can help teams:

  • Identify unexpected connections. A workload that starts contacting an unfamiliar server or service may merit review.
  • Investigate lateral movement. Connections between hosts can provide context about how an incident progressed.
  • Find exposed or unused services. Observed communication can help validate firewall rules and segmentation plans.
  • Detect unusual data movement. Changes in connection volume or destination can be useful signals, though they require context.
  • Support troubleshooting. Network evidence can help distinguish an application issue from a connectivity or policy issue.
  • Validate security controls. Teams can compare observed communication with intended access rules and investigate exceptions.

Monitoring does not automatically prevent an attack or prove that a connection is malicious. It produces evidence that needs to be interpreted alongside asset inventories, identity data, application context, and other security telemetry.

Approaches to east-west traffic monitoring

No single collection method provides complete visibility in every environment. The main options can be combined, but their tradeoffs should be understood before deployment.

1. Network taps and packet brokers

A network tap copies packets from a physical link to a monitoring system. Packet brokers can aggregate, filter, or distribute that traffic to sensors. This approach is useful in environments where teams control the physical network and need packet-level evidence.

Strengths

  • Can provide detailed packet data, subject to the placement and capabilities of the sensor.
  • Does not usually require installing software on each monitored endpoint.
  • Can support multiple analysis tools from shared collection points.

Tradeoffs

  • Visibility depends on where taps are placed. A sensor on one link cannot see traffic that never crosses it.
  • High traffic volumes can exceed capture or processing capacity, resulting in dropped packets or incomplete records.
  • Physical access, cabling, and changes to network topology can make deployment more involved.
  • Encryption limits what a passive sensor can learn from application payloads, although metadata may remain available.

Taps are a strong option for important network chokepoints, but they are not a guarantee of full internal coverage. Document what each sensor can and cannot see.

2. Switch port mirroring

Switch port mirroring, sometimes called SPAN, copies selected switch traffic to a monitoring port. It is often easier to configure than a physical tap and can be practical for focused monitoring or testing.

Strengths

  • Uses existing switching equipment.
  • Can be configured to mirror selected ports or virtual networks, depending on the switch.
  • Offers a relatively accessible way to feed packets to a sensor.

Tradeoffs

  • A mirror destination may receive more traffic than it can handle, especially when several high-volume ports are copied to one interface.
  • Some switching configurations can drop mirrored packets under load.
  • The mirrored traffic may not include every path, especially where traffic stays within a host, virtual switch, or cloud fabric.
  • Configuration can change as ports, workloads, and network layouts change.

Treat a SPAN feed as a source with known capacity limits. Validate that it is capturing the traffic you expect rather than assuming that a configured mirror equals complete visibility.

3. Flow records and network telemetry

Flow records summarize conversations rather than preserving every packet. Depending on the exporter and format, records can include source and destination addresses, ports, protocol, timestamps, and byte or packet counts. Common technologies include IPFIX and vendor-specific flow formats.

Strengths

  • Usually requires less storage and processing than full packet capture.
  • Can provide broad visibility into communication patterns and traffic volume.
  • Useful for baselining, identifying unusual peers, and investigating when a connection occurred.

Tradeoffs

  • Summaries do not provide the full packet sequence or application payload.
  • Export sampling, timeouts, and device support affect the detail and accuracy of records.
  • Flow data alone may not identify the user, process, or workload responsible for a connection.
  • High-cardinality environments can still generate substantial data and analysis costs.

Flow telemetry is often a sensible foundation for broad coverage. It can be paired with more detailed packet or endpoint evidence for high-risk systems and investigations.

4. Host-based endpoint telemetry

Endpoint agents and operating-system tools can observe connections from the host itself. Depending on the platform and configuration, they may associate a connection with a process, user, or workload identity—context that network-only records may lack.

Strengths

  • Can see activity that remains within a host, such as communication between local processes or containers, if the sensor supports it.
  • May provide process and identity context for a network connection.
  • Can follow workloads across network changes when the agent remains installed and reporting.

Tradeoffs

  • Requires deployment, maintenance, and health monitoring across supported systems.
  • Coverage can be affected by unsupported platforms, agent failures, or tampering.
  • Host telemetry may add operational overhead and needs careful tuning.
  • Data collection must align with privacy, employee-monitoring, and retention requirements.

Endpoint telemetry is valuable when attribution matters, but it should not be assumed to replace network visibility. A layered design can use host evidence to explain connections that flow records first reveal.

5. Cloud and virtual-network telemetry

Public cloud platforms and virtualization systems may provide flow logs, firewall logs, virtual network mirroring, or service-specific audit records. These sources can offer visibility into traffic that does not pass through an organization’s physical network sensors.

Strengths

  • Can cover cloud network paths and resources that are difficult to observe with on-premises taps.
  • Often integrates with cloud resource metadata, helping teams associate traffic with virtual machines, interfaces, or subnets.
  • Can be enabled or scoped through infrastructure configuration, subject to platform capabilities.

Tradeoffs

  • Coverage, fields, retention, and delivery behavior differ by provider and service.
  • Logging may incur storage, processing, or data-transfer costs.
  • A flow record is not necessarily real-time, complete, or equivalent to packet capture.
  • Resource identifiers and network layouts can change quickly in dynamic environments.

Review each provider’s documentation and test the actual records generated in your account. Confirm the scope, delivery delay, supported traffic types, and retention settings before relying on the data for detection or incident response.

Choosing an approach

A practical design starts with the questions the monitoring program must answer, not with a tool purchase. Consider these criteria:

| Criterion | Questions to ask |

|---|---|

| Coverage | Which data centers, cloud networks, clusters, hosts, and network paths must be visible? What traffic bypasses the proposed collection points? |

| Required detail | Do investigators need connection summaries, process attribution, packet contents, or only alerts about policy violations? |

| Scale | What are normal and peak traffic volumes? Can sensors, links, and storage handle them without unacceptable loss? |

| Latency | Is near-real-time alerting required, or is delayed analysis sufficient for some use cases? |

| Encryption | What metadata remains observable, and where can authorized application or endpoint telemetry provide additional context? |

| Operations | Who will maintain sensors, parsing, detections, and data pipelines? How will collection failures be detected? |

| Cost and retention | What data volume will be stored, for how long, and at what cost? Can high-value evidence be retained longer than routine telemetry? |

| Privacy and governance | Is collection proportionate, access controlled, and consistent with applicable policies and legal requirements? |

For many organizations, a layered design is more practical than trying to collect full packet data everywhere. For example, broad flow records can establish normal communication patterns, while endpoint telemetry or packet sensors provide deeper evidence for critical systems, sensitive segments, or investigations. Cloud-native sources can fill gaps in environments that physical network sensors cannot reach.

A practical deployment plan

1. Define priority questions and assets

Start with concrete use cases: detecting unexpected database access, investigating connections to identity infrastructure, or finding workloads that communicate across a restricted boundary. Map the systems and traffic paths relevant to those questions. Include east-west paths within hosts, virtual networks, and cloud environments where applicable.

2. Document collection coverage

For each data source, record its location, traffic scope, fields, sampling behavior, retention, and known blind spots. A diagram or inventory should make it clear which paths are visible and which are not. This prevents teams from mistaking partial monitoring for comprehensive coverage.

3. Establish a baseline

Collect enough representative telemetry to understand ordinary communication patterns, including expected maintenance windows and deployment changes. Group activity by useful context—such as service, environment, subnet, workload, or business owner—rather than relying only on raw IP addresses. Baselines should be reviewed as infrastructure changes.

4. Build detections around meaningful context

Start with a small number of high-value detections. Examples include a production application contacting an unexpected administrative service, a server making new connections to many peers, or a workload communicating with a restricted data store outside its expected role. These are investigation leads, not automatic proof of compromise.

Enrich alerts with asset ownership, environment, identity, and change-management context where available. Tune detections to reduce routine noise while preserving visibility into meaningful exceptions.

5. Validate data quality and response

Test whether expected traffic appears in the monitoring platform and whether known gaps are understood. Check timestamp consistency, field parsing, sensor health, dropped-packet indicators, and log delivery. Run tabletop or controlled validation exercises to confirm that analysts can find and interpret the evidence they need.

6. Review and adjust

Reassess coverage when networks, cloud accounts, applications, or segmentation rules change. Monitor the monitoring system itself: a silent sensor or broken export pipeline can create a blind spot that looks like an absence of suspicious traffic.

Security and operational best practices

  • Use least privilege for telemetry systems. Restrict who can view, export, and administer network data, which may reveal sensitive infrastructure details.
  • Protect collection and storage paths. Apply appropriate authentication, encryption in transit and at rest, and access logging.
  • Set retention intentionally. Retain enough evidence for operational and incident-response needs while respecting privacy and data-minimization requirements.
  • Separate observation from enforcement. Monitoring identifies activity; firewalls, access controls, and segmentation policies restrict it. Coordinate the two, but do not treat one as a substitute for the other.
  • Plan for encrypted traffic. Do not assume passive monitoring can inspect encrypted payloads. Use permitted metadata, endpoint context, application logs, or authorized decryption architectures where appropriate.
  • Track exceptions. Temporary rules and one-off connections can become permanent blind spots. Assign owners and review dates to exceptions.
  • Measure coverage and quality, not just alert counts. Useful measures include monitored asset coverage, data-source health, known blind spots, and time to locate relevant evidence during an investigation.

Common limitations and mistakes

A frequent mistake is deploying a sensor at a convenient location and assuming it can see all internal communication. Traffic may take other routes, stay within a host, or traverse a cloud fabric outside the sensor’s scope. Another is treating high packet volume as proof of effective monitoring: a feed can be overwhelmed, sampled, or poorly parsed.

Teams can also create noisy alerts by labeling any unusual connection as malicious. New deployments, backups, service discovery, and administrative work can all change traffic patterns. Detection logic should incorporate business context and provide enough evidence for analysts to decide what to do next.

Finally, monitoring data is not a replacement for sound architecture. Identity controls, patching, endpoint security, secure configuration, and network segmentation remain important defenses. Monitoring can reveal when expected boundaries or behaviors are changing; it cannot make weak access controls safe by itself.

FAQ

Is east-west traffic monitoring the same as network segmentation?

No. Segmentation uses network or identity controls to limit which systems can communicate. Monitoring observes communication and helps identify unexpected activity or validate whether controls behave as intended. They complement one another.

Can encrypted traffic still be monitored?

Yes, but the available detail depends on the collection method. Sensors may record connection metadata such as endpoints, timing, and volume without seeing encrypted application contents. Endpoint, application, or identity telemetry can add context. Decryption should be considered only when justified and properly authorized.

Should an organization capture packets everywhere?

Not necessarily. Full packet capture can provide valuable detail, but it has storage, processing, privacy, and operational costs, and may still miss traffic outside sensor coverage. Flow records or endpoint telemetry may provide more practical broad coverage, with packet capture reserved for selected locations or investigations.

How do teams know whether monitoring coverage is sufficient?

Define the assets and paths relevant to your security objectives, then map each to a functioning data source. Validate that expected connections appear, identify unmonitored paths, and track sensor and pipeline health. Coverage is an ongoing property of the architecture, not a one-time configuration task.

Conclusion

East-west traffic monitoring gives teams evidence about how internal systems communicate, helping them investigate incidents, spot unexpected connections, and validate security controls. Network taps and mirrored ports offer packet visibility at selected points; flow records provide scalable connection summaries; endpoint telemetry can add process context; and cloud-native logs can cover virtual networks that physical sensors miss.

The right design depends on the questions you need to answer, the paths you can observe, and the detail you can operate responsibly. Start with critical assets and explicit use cases, document blind spots, validate data quality, and combine monitoring with segmentation and access controls. A measured, layered approach is usually more useful than collecting everything without a clear plan.