Lab N3: Security Intelligence — DFW Rule Recommendation
Objectives
- Understand where Security Intelligence runs: a feature of the NSX Application Platform (Kubernetes), licensed with vDefend Firewall
- Activate Security Intelligence and flow collection for a defined application scope
- Use the recommendation engine to generate DFW groups and rules from observed flows
- Choose the recommendation inputs deliberately: time range, traffic type, connectivity strategy, group reuse threshold, infrastructure exclusion
- Publish recommendations with an allow-everything-else (Deny list) strategy and logging first, then move to Allow list (default drop)
- Read the Discover & Take Action canvas to validate the observed application topology
Prerequisites
VCF 9.0 workload domain with NSX Manager, an NSX Application Platform instance deployed (3 control-plane + at least 4 worker nodes, on a Supervisor-managed Kubernetes cluster) with Security Intelligence activated, vDefend Firewall licence applied, and at least 3 VMs forming a multi-tier application with active traffic between tiers
Prior labs: net-virt-01, net-virt-02, net-virt-03
Required skills:
- DFW policy concepts (from Lab net-virt-03)
- Application flow pattern understanding
- Micro-segmentation principles
Lab Environment
VCF 9.0 workload domain NSX Manager cluster paired 1:1 with an NSX Application Platform instance (recommended to run on the management domain Supervisor). A 3-tier application (web-app-db) running with real traffic patterns: web→app on HTTP/HTTPS, app→db on MySQL/PostgreSQL, web→external for client access. Security Intelligence keeps flow data for the recommendation time ranges (last hour, 24 hours, 1 week, 1 month).
Tasks
Task 1 Flow-Based Micro-Segmentation with Security Intelligence
Use Security Intelligence flow telemetry and its recommendation engine to discover application communication patterns and generate DFW groups and rules — replacing the manual rule-writing process from Lab net-virt-03 with an observation-driven approach, then staging enforcement through the connectivity strategy. Observe first, then enforce.
Verify the platform. NSX Manager → System → NSX Application Platform: the platform shows Up and the Security Intelligence feature is Activated (activate it from the Features card if not). Confirm the vDefend Firewall licence is applied (Security Intelligence is part of that add-on). Platform logs for later troubleshooting live in the NAPP support bundle (/var/log/napp/supportbundle/<worker>/…), not on an appliance.
Define the application scope and data collection. Create or confirm an NSX group with every member of the application you are securing (SEC-VIWLD-017), here the VMs tagged env=production from Lab net-virt-03. Confirm data collection is active for the clusters (or standalone hosts) running those VMs: NSX Manager → System → Settings → NSX Intelligence → Data Collection tab, per the Security Intelligence 4.2 docs (record the label your build shows). Collection starts on all hosts and clusters by default after activation and can only be changed per cluster, not per host in a cluster. Flows are then recorded per vNIC: source, destination, port, protocol, bytes.
Generate application traffic (if not already flowing). Simulate real application patterns: (a) From web VMs, run repeated HTTP requests to app VMs: while true; do curl -s http://10.10.2.11; sleep 1; done. (b) From app VMs, run MySQL queries to db VMs: mysql -h 10.10.3.11 -u app -p -e 'SELECT 1'. (c) From web VMs, generate DNS lookups and NTP sync. Let traffic flow for at least 1 hour (24-48 hours in production) to build a representative baseline.
Explore traffic visualization. NSX Manager → Plan & Troubleshoot → Discover & Take Action. In the Security view selector choose Groups or Computes; the canvas shows entities as nodes and flows as edges, with a Flows bar for the time range. Look for: (a) tightly coupled groups — application tiers; (b) unexpected connections — lateral movement or misconfigured services; (c) high-volume flows — database connections, backup traffic. Compare the observed topology with the documented architecture.
Generate a recommendation. Plan & Troubleshoot → Recommendations → Start New Recommendation (or right-click the application group on the Discover & Take Action canvas → Start Recommendation). Set: Recommendation Name; input entities = the application group; Time Range (Last 1 hour / 24 hours / 1 week / 1 month — pick the window that includes the application's periodic jobs, SEC-VIWLD-019); Traffic Type (e.g. Incoming and Intra-application Traffic, SEC-VIWLD-020); Connectivity Strategy for the default rule (Allow list → drop unmatched, Deny list → allow unmatched, SEC-VIWLD-018); group reuse threshold (10–100 %); toggle to exclude infrastructure compute entities. Click Start Discovery. The job moves from Discovery in Progress to Ready to Publish.
Review recommendation quality. Expand the Ready to Publish recommendation: it lists the proposed groups (reused existing groups above the reuse threshold, new ones below it), services and rules, and the canvas icon shows the entities and flows it was built from. Prune before publishing: remove rules created by management tools that ran during observation (vulnerability scanners, backup agents that reach every VM) or exclude them via the infrastructure classification; add known flows that did not occur in the time range (monthly batch jobs); decide where an L7 (APP-ID) rule is worth the DPI cost (SEC-VIWLD-021).
Publish in observation mode. For the first publication choose Connectivity Strategy = Deny list (the default rule allows unmatched traffic) and enable logging on the recommended rules, so the policy exists in the Application category but nothing legitimate is dropped. Watch the DFW rule hit counters and logs, and the Discover & Take Action canvas (Flows bar set to the period after publication), for flows that hit the default rule; each such flow needs an allow rule or a conscious decision to block it. Turn on the recommendation's Monitor toggle so scope or strategy changes are checked hourly.
Iterate and refine. From the observation period: (a) add allow rules for legitimate flows that hit the default rule; (b) remove allow rules for one-time or test traffic that should not persist; (c) confirm management-plane traffic (vCenter, NSX Manager, monitoring agents) is covered by Infrastructure-category rules created outside the recommendation (SEC-VIWLD-022). Rerun the recommendation (the Monitor toggle flags scope or strategy changes) and compare. Two or three rounds are typical before the policy is ready for Allow list.
Switch to enforcement. Once the default rule shows no legitimate hits, rerun the recommendation with Connectivity Strategy = Allow list (default rule drops unmatched traffic) and publish, or edit the published default rule from Allow to Drop. Keep logging on for the first day and a rollback ready (set the default rule back to Allow). Record the decision in the design: scope, time range, strategy and L4/L7 choice (SEC-VIWLD-017..021).
DFW Rule Analysis (post-enforcement optimization). After enforcement, use Rule Analysis in the vDefend Security Services Platform (SSP 5.1, Understanding Rule Analysis and Initiate Rule Analysis): activate it on the Rule Analysis feature card (System → Platform & Features), run an analysis on demand (Go to Rule Analysis → Start Analysis) or turn on Auto Analysis with a weekly, monthly or quarterly frequency, and review the results on the Rule Analysis dashboard in the NSX Manager UI (download the report first: a new run replaces it). It reports seven anomaly categories: (a) Duplication — a rule duplicates another rule; (b) Contradiction — two rules have the same configuration but different actions; (c) Redundancy — a lower-precedence rule whose traffic is a subset of a higher rule with the same action; (d) Shadowing — a lower-precedence rule that never takes effect because a higher rule with a different action already matches its traffic; (e) Consolidation — two rules share the same source, destination, applied-to fields and action but differ only in services or context profiles, so they can be merged (provided no intervening rule is affected); (f) Permissiveness — a rule that allows all traffic from ANY source to ANY destination using ANY service; (g) Irrelevance — a rule that references resources that do not exist or cannot be applied, such as an empty group or an IP-set-based group in the Applied-To field of a DFW rule (category definitions: SSP 5.1 Understanding Rule Analysis; VMware Validated Solutions 9.x p.251). Rules with no hits are not an analysis category — find them from the DFW rule hit counters. Act on these findings to keep the rule set lean.
Validation Gate
Check: Flow-based micro-segmentation rules generated, validated during an observation period (Deny list with logging), and enforced
Expected: Security Intelligence collecting flows from all target VMs, Discover & Take Action canvas showing the correct application topology, recommended rules covering all legitimate flows, observation period shows no legitimate default-rule hits, Allow list (default drop) published with monitoring
Common Errors
Final Validation
Automated micro-segmentation operational with flow-based rule generation and optimization
✓ Flow collection status → NSX Application Platform Up, Security Intelligence Activated, flows visible for all target VMs
✓ Traffic visualization → Application topology correctly mapped in flow visualization dashboard
✓ Rule recommendations → Generated rules cover all observed legitimate traffic patterns
✓ Observation period validation → No legitimate flows hitting the logged default rule after the observation window
✓ Enforcement mode → All rules enforced with no connectivity disruptions
✓ Rule analysis → Rule Analysis findings reviewed and acted upon (duplication, redundancy, shadowing and consolidation anomalies resolved); zero-hit rules reviewed from the DFW hit counters
Cleanup / Restore
• Set the published default rule back to Allow or delete the generated policy
• Disable data collection for the lab VMs to limit NSX Application Platform flow storage
• Delete the recommendation job if it is no longer needed
Design Reflection (VCDX)
Observation-driven micro-segmentation is a strong VCDX story because it shows you understand that manual rule creation does not scale. Defence points: (1) rules come from observed flows rather than documentation alone; (2) the connectivity strategy stages enforcement — Deny list (default allow, logged) before Allow list (default drop) — so there is no big-bang cut-over; (3) scope, time range and traffic type are explicit design decisions (SEC-VIWLD-017..020); (4) Security Intelligence is an NSX Application Platform feature, so platform sizing (3 control-plane + ≥4 worker nodes, 100 % reserved) and its 1:1 pairing with an NSX Manager cluster belong in the design; (5) emergency, infrastructure and environment rules stay outside the recommendations (SEC-VIWLD-022).
Requirements
- Micro-segmentation for 100+ VMs across multiple applications
- Minimal risk of blocking legitimate traffic during enforcement
- Ongoing rule optimization as application patterns evolve
Constraints
- Security Intelligence requires an NSX Application Platform instance (Kubernetes) paired 1:1 with the NSX Manager cluster; worker-node count scales with flow volume and retention
- Observation window must capture all traffic patterns including periodic jobs
- Recommendation quality depends on representative traffic during the chosen time range
Assumptions
- Applications generate representative traffic patterns within 48-hour observation window
- Management and monitoring traffic is identified and excluded from application-level recommendations
- Operations team commits to the staged Deny list → Allow list methodology (not big-bang)
Risks
- Short time range misses periodic traffic → enforcement blocks legitimate flows
- Under-sized NSX Application Platform drops flows → recommendations based on partial data
- Rule sprawl over time as new recommendations are published without retiring stale rules
Self-Assessment Discussion Prompts
- How do you handle micro-segmentation for applications with seasonal traffic patterns (e.g., month-end batch jobs)?
- What is the recommended observation window for a 500-VM environment with mixed workloads?
- How does Security Intelligence handle encrypted traffic (TLS) — what can it still recommend, and when do L7 (APP-ID) rules add value (SEC-VIWLD-021)?
- Compare recommendation-driven segmentation with the manual approach of Lab net-virt-03 — when is each appropriate, and which rule categories must stay manual?
Extensions
Use the NSX API to export Security Intelligence recommendations as JSON and integrate with a CI/CD pipeline for policy-as-code
Use Security Intelligence suspicious-traffic detections (sent to NDR when activated) for anomalous traffic patterns — flows that don't match any existing DFW rule
Build a quarterly rule hygiene workflow using Rule Analysis to prune shadowed and redundant rules, and the DFW rule hit counters to find unused rules
Correlate Security Intelligence flow data with VCF Operations (Aria Operations) for cross-domain performance correlation
⚠ Known Pitfalls
References
- VCF 9.0 documentation — NSX Application Platform and Security Intelligence: https://techdocs.broadcom.com/us/en/vmware-cis/vcf/vcf-9-0-and-later/9-0.html
- VMware Validated Solutions 1.0 — Lateral Security for VCF with VMware vDefend (Security Intelligence, SEC-VIWLD-016..022)
- vDefend Security Intelligence 4.2 — Generate a New Security Intelligence Recommendation (techdocs.broadcom.com)
- KB 373793 — NSX Application Platform sizing tool