1. Intro2. Key Insight3. Timeline4. Section 15. Challenges6. Analysis7. Reference8. Section 29. Roadmap10. Closing Thoughts11. Section 312. Section
READ:
OT/ICS Fundamentals 04-AUG-2026 · 12 min read

Purdue is Dead: Brownfield OT Segmentation

Strict Purdue levels create operational friction in brownfield OT. This technical deep-dive examines real-world limitations and details proven alternatives including unidirectional gateways, protocol-aware microsegmentation at Level 2, data diodes, and zero-trust flow policies that deliver defensible architectures without greenfield replacement.

Network segmentationBrownfield OTData diodesZero Trust OTS4 discussion
PractitionerAdvancedPR.DS Data SecuritySecurity Hardening
Article Details
CategoryOT/ICS Fundamentals
Published04-AUG-2026
Read Time12 min read
AuthorNEXUS Engineering
The Problem

The Purdue Model Has Reached End of Life in Brownfield OT

"Purdue was never designed as a security boundary model yet became the default reference architecture."

Purdue is dead. The hierarchical model formalized in ISA-95 and widely adopted via Purdue Reference Model defined strict functional levels from Level 0 field devices through Level 5 enterprise systems. In brownfield OT deployments this creates unsustainable tension. Legacy systems installed before 2010 rarely support the authentication primitives required for strict conduit enforcement. Control engineers routinely bypass firewalls to maintain sub-100ms polling cycles on DNP3 and Modbus TCP networks. The result is implicit trust across what should be isolated zones. Examination of the 2017 TRITON incident reveals precisely this failure mode where adversaries traversed from compromised Level 3 engineering stations directly to Level 1 safety controllers because conduit rules existed only on paper.

Legacy brownfield constraints compound the architectural mismatch. Many Level 1 PLCs still operate on proprietary protocols lacking sequence numbers or cryptographic integrity checks. Introducing Layer-3 segmentation at these points introduces jitter that violates process safety time requirements in SIS loops. Industry observations from S4 conferences document repeated cases where segmentation projects were rolled back after causing nuisance trips in continuous processes. NIST SP 800-82 Revision 3 acknowledges these realities by shifting emphasis toward risk-based zoning rather than dogmatic Purdue adherence. Yet many compliance-driven programs still treat the five-level diagram as immutable, generating policy that cannot be implemented without halting production.

IIoT convergence has rendered the classic pyramid obsolete. Modern brownfield sites stream telemetry from Level 0 sensors directly to cloud analytics platforms via edge gateways that straddle multiple legacy levels. Historians at Level 2 pull data from both safety and basic process control networks without granular flow control. Remote access solutions for OEM support create dynamic conduits that static ACLs cannot govern. The 2021 Oldsmar water treatment incident demonstrated how an adversary with compromised IT credentials could reach Level 0 HMIs because no explicit positive control existed on the intermediate conduits. These patterns repeat across energy, manufacturing and transportation sectors according to Dragos Year in Review reports.

Quantitative data underscores the gap. Analysis of MITRE ATT&CK for ICS telemetry shows that 68 percent of observed lateral movement techniques exploit poorly defined or unenforced conduits between Purdue levels. IEC 62443-3-2 risk assessment methodology explicitly calls for zone and conduit modeling yet brownfield implementations frequently map zones directly to Purdue levels without accounting for exception flows. The outcome is a false sense of segmentation that sophisticated adversaries bypass using living-off-the-land techniques within trusted maintenance pathways. Practitioners therefore require alternatives that prioritize enforceable data flows over topological purity.

The counter-argument is straightforward. Insisting on classic Purdue segmentation in brownfield sites consumes engineering resources while delivering marginal risk reduction. A pragmatic shift toward flow-centric policies, hardware-enforced unidirectional links and asset-specific microsegmentation yields higher security outcomes with lower operational disruption. This article maps those practical patterns with concrete implementation details drawn from operational deployments.

⚡

Classic Purdue levels create false security boundaries in brownfield OT. Replace them with enforceable zero-trust data flows, hardware data diodes enforcing physical one-way transfer, and protocol-aware microsegmentation at Level 2 to reduce lateral movement without violating deterministic timing.

Historical Context

Evolution Beyond Purdue in OT Security

1995
Purdue Model Formalized

ISA-95 creates hierarchical reference architecture later adopted as de-facto segmentation blueprint despite lacking security design intent.

2015
Industroyer Attack

Malware demonstrates protocol manipulation across flat Level 2 networks highlighting conduit weaknesses in Ukrainian substations.

2017
TRITON/TRISIS Incident

Safety system compromise via Level 3 to Level 1 traversal proves Purdue segmentation insufficient against determined adversaries.

2021
Oldsmar Incident

Remote access bypass of all logical controls shows legacy Purdue assumptions enable direct Level 0 manipulation.

2023
NIST SP 800-82r3

Updated guidance prioritizes dynamic zoning and data-flow policies over static Purdue levels for modern OT environments.

Practical Alternatives

What Actually Works: Unidirectional Gateways and Microsegmentation

Unidirectional gateways provide the strongest enforcement mechanism available to brownfield operators. Hardware data diodes from vendors such as Waterfall Security and Owl Cyber Defense use physical layer optical isolation ensuring zero return path exists even if the receiving host is fully compromised. In practice these devices replicate historian data, SCADA metrics and IIoT telemetry from Level 1/2 into Level 3/4 networks without exposing OT assets to inbound commands. Deployment behind existing Level 2 switches allows selective replication of OPC UA packets while blocking all TCP handshakes in the reverse direction. Measured latency impact remains under 50ms for most historian use cases making the technology compatible with near-real-time control oversight.

Microsegmentation within Level 2 delivers granular policy enforcement where Purdue hierarchies fail. Modern OT-aware platforms such as Claroty CTD and Nozomi Networks Guardian map every asset by MAC, firmware version and protocol fingerprint then enforce allow-list policies on specific command codes. For example a policy may permit only Function Code 3 reads from a specific HMI to a PLC while blocking Function Code 6 writes except during authenticated maintenance windows. This approach respects deterministic timing because enforcement occurs at wire speed within the switch fabric using eBPF or dedicated DPI appliances. Brownfield integration leverages passive monitoring taps to baseline flows before shifting to enforcement mode, minimizing surprise outages.

Zero-trust data flow policies complement the above technologies by adding contextual authorization. Rather than relying on network location, policies evaluate device identity, user role, firmware integrity, and process state before allowing any transaction. Integration with existing IAM systems allows service accounts tied to specific maintenance tasks to open temporary conduits lasting less than 15 minutes. In brownfield environments this requires careful mapping of existing exception flows documented during IEC 62443-3-2 zone workshops. The non-obvious insight is that combining data diodes for bulk telemetry with microsegmentation for command-and-control paths and zero-trust policy engines for human-initiated flows creates defense-in-depth without requiring forklift upgrades to legacy controllers.

Implementation data from critical infrastructure operators shows 92 percent reduction in viable lateral movement paths when these three technologies are layered. The key is maintaining strict positive control: every flow must be explicitly authorized rather than implicitly permitted by sitting inside a Purdue level. Protocol-aware inspection of S7comm, CIP, DNP3, Modbus and OPC UA allows policy granularity down to individual function codes and register blocks. This level of precision was unavailable when the Purdue model was created yet is now table stakes for mature OT security programs.

Integration with existing Purdue documentation remains straightforward. Legacy diagrams are retained for functional reference while security architecture overlays define zones based on data sensitivity and impact rather than level number. Conduits become explicit policy bundles enforced at multiple control points. This hybrid representation satisfies both operational staff familiar with Purdue and auditors requiring evidence of IEC 62443 conformance.

Implementation Reality

Key Brownfield Segmentation Challenges

Retrofitting segmentation into live OT environments introduces constraints rarely discussed in greenfield design guides. The following challenges reflect observed patterns across energy, chemical and discrete manufacturing sites pursuing post-Purdue architectures.

critical

Deterministic Timing Conflicts

Introduction of inspection or policy enforcement points can add jitter exceeding process safety margins in sub-50ms control loops. Legacy PLC scan cycles and safety instrumented systems frequently reject even 10ms added latency, forcing security teams to carve out exceptions that reintroduce implicit trust. Mitigation requires hardware-accelerated DPI appliances placed in parallel with data diodes.

high

Legacy Protocol Opacity

Many brownfield devices use undocumented or proprietary extensions to Modbus, Profibus and DeviceNet that defeat commercial DPI signatures. Without accurate asset inventory and protocol baseline captured via passive monitoring, microsegmentation policies risk blocking legitimate vendor-specific diagnostic commands required during maintenance windows.

medium

Change Management Windows

OT maintenance occurs during scheduled outages measured in hours per year. Deploying segmentation policies requires exhaustive pre-staging, offline testing on digital twins and rollback plans. Organizations lacking mature digital twin capabilities face extended timelines measured in quarters rather than weeks.

Purdue Legacy vs Adaptive Segmentation

What Works

  • Hardware data diodes provide physical enforcement with zero return path
  • Microsegmentation at Level 2 respects deterministic timing using DPI at wire speed
  • Contextual zero-trust policies evaluate device health before authorizing flows
  • Passive monitoring baselines enable safe transition from detection to enforcement
  • IEC 62443-3-2 zoning maps cleanly to flow-centric conduits

What Doesn't

  • Static ACLs on Level 3 firewalls cannot inspect OT protocol semantics
  • Assuming Purdue levels equal security zones creates implicit trust
  • Retrofitting Layer 3 routing into Level 1 introduces unacceptable jitter
  • Compliance checklists demanding strict Purdue adherence ignore brownfield physics
  • Vendor remote access VPNs routinely bypass all defined conduits
Technology Comparison

Segmentation Technologies for Brownfield OT

ApproachEnforcement TypeBrownfield CompatibilityLatency ImpactPrimary Use CaseExample Vendors
Classic Purdue LevelsLogical hierarchyLowNoneFunctional reference onlyN/A
Traditional FirewallsLayer 3/4 ACLsMedium2-5msIT/OT boundaryCiscoFortinet
Data DiodesPhysical one-wayHigh<1msTelemetry exportWaterfallOwl
Protocol MicrosegmentationDPI policyHigh0.2-1msLevel 2 east-westClarotyNozomi
Zero-Trust Policy EngineContextual authMedium1-3msHuman and app flowsPalo AltoIllumio
Outcomes

Measured Results from Post-Purdue Deployments

Organizations that have moved beyond Purdue report concrete improvements. One North American pipeline operator replaced traditional Level 3 firewalls with a combination of data diodes for historian replication and microsegmentation within the control center Level 2 network. The project reduced the number of assets visible to the enterprise network from 340 to 17 while maintaining full SCADA visibility. Attack surface metrics generated by their Nozomi deployment showed 87 percent reduction in exposed OT protocols. Importantly the project completed without any unscheduled downtime during a six-month phased rollout.

A European chemical manufacturer faced repeated audit findings for flat network architecture between batch control and safety systems. Rather than attempting full Purdue realignment they deployed unidirectional gateways to carry safety status data to the MES layer and implemented zero-trust policies on all engineering workstation access. The solution satisfied IEC 62443-3-2 clause 4.2 zone requirements by treating safety systems as a separate high-impact zone with strictly controlled conduits. Post-deployment red team exercises could not reach SIS logic from the business network, a result that had eluded previous Purdue-based projects.

These outcomes share common success factors. First, passive asset discovery and flow mapping preceded any enforcement. Second, security teams collaborated closely with process engineers to validate timing budgets before policy activation. Third, digital twins or offline test beds were used to rehearsing failure modes. The non-obvious insight from these deployments is that segmentation effectiveness correlates more strongly with quality of flow baseline than with depth of Purdue documentation.

Looking forward, convergence of OT security platforms with IT zero-trust frameworks will further simplify brownfield segmentation. However the physics of legacy control systems will continue to require hybrid solutions blending hardware isolation with intelligent policy enforcement. Practitioners should therefore focus on measurable risk reduction metrics such as mean time to detect lateral movement rather than compliance checkboxes tied to outdated architectural models.

The roadmap below translates these lessons into actionable phases suitable for most brownfield sites.

Practical Path Forward

Implementation Roadmap for Brownfield Segmentation

▲

All phases assume existing passive monitoring capability is in place before enforcement begins.

Phase 1
Months 1-3

Discovery & Baseline

Conduct full asset inventory, protocol dissection and flow mapping across current Purdue levels. Produce zone and conduit model per IEC 62443-3-2 that reflects actual data flows rather than theoretical hierarchy. Identify high-risk implicit trust paths for immediate attention.

Passive sensor deploymentAsset fingerprintingFlow baseliningRisk ranking of conduits
Phase 2
Months 4-6

Unidirectional Telemetry

Deploy data diodes for all outbound historian, monitoring and backup flows. Configure replication services to maintain continuity for MES and cloud analytics. Validate one-way enforcement using red team simulation attempting command injection through the diode.

Hardware installationReplication configurationFailover testingPerformance benchmarking
Phase 3
Months 7-12

Microsegmentation & Zero-Trust

Implement policy enforcement at Level 2 using OT-aware DPI. Roll out contextual zero-trust authorization for all interactive and machine-to-machine sessions. Integrate with existing IAM and maintenance ticketing systems to automate temporary conduit activation.

Policy authoringPilot enforcementRed team validationOperator training
Closing Thoughts

Questions Worth Sitting With

The shift away from Purdue-centric thinking requires honest assessment of operational constraints versus security outcomes.

01

How do you objectively measure lateral movement risk in your current brownfield conduit design?

02

When was the last time your segmentation policies were validated against actual OT protocol flows rather than IP subnets?

03

What would a zero-trust policy for your most critical safety instrumented system look like in practice?

04

How do you balance the need for rapid vendor remote access against the requirement for positive control on all conduits?

05

Which of your current Purdue level assumptions would collapse first under a living-off-the-land adversary?

Effective OT segmentation is no longer about drawing clean hierarchical boxes. It is about enforcing explicit, verifiable, context-aware data flows at every boundary.

Conclusion: Beyond Purdue Lies Resilience

First paragraph of content.

Second paragraph continues the narrative.

← Back to CyberCuriosity Let's Talk
Comments & Suggestions
Thoughts on this article? Corrections, questions, or additions — all welcome.
★ ★ ★ ★ ★ Optional — tap to rate
GDPR: Your data is processed solely to respond to your enquiry and is never shared with third parties. By submitting you consent to NEXUS Cybersecurity storing your details for this purpose only.
Sent privately — never published publicly