Pourquoi la micro-segmentation est-elle cruciale pour le Zero Trust dans les ICS
Découvrez pourquoi la micro-segmentation est essentielle pour mettre en œuvre une sécurité Zero Trust dans les systèmes de contrôle industriel, réduisant les risques et renforçant la résilience du réseau OT.
📖 Temps de lecture estimé : 5 minutes
Article
Why Microsegmentation is Critical for Zero Trust in ICS
Industrial Control Systems (ICS) sit at the heart of critical infrastructure: power grids, water treatment plants, oil refineries, and manufacturing facilities. Their once air-gapped networks are increasingly converging with IT systems in pursuit of operational agility—opening long-standing proprietary environments to external connectivity and, with it, new adversarial risks. As the threat landscape expands, reliance on trusted network perimeters has proven fatally insufficient, yielding way to the Zero Trust security paradigm. Within this context, microsegmentation emerges as a technical linchpin, not just for theoretical rigor but for practical, risk-reducing architecture in real-world OT settings.
Historical Context: Legacy ICS Network Architectures and Their Assumptions
Traditional ICS/OT networks were architected on the premise of implicit trust—if you were inside the network, you were presumed trustworthy. Flat Layer 2/3 topologies, driven by deterministic protocols like Modbus, DNP3, and PROFINET, facilitated broad device-to-device communication for operational efficiency. At best, early segmentation efforts involved only basic VLANs or physical separation, enforced by manually-configured routers, firewalls, or even air gaps.
In the evolution of IT networking, segmentation dates back at least to the late 1980s when VLANs (IEEE 802.1Q) were introduced to logically separate broadcast domains. Firewalls entered the picture to control north-south traffic (ingress/egress), eventually supporting stateful inspection and rule-based controls. However, east-west (lateral) movement inside trusted zones was rarely, if ever, constrained—especially in ICS environments where operational simplicity and deterministic performance trumped all else.
Zero Trust in ICS: Principles and Challenges
Zero Trust, a term coined by Forrester’s John Kindervag in 2010, discards the notion of a trusted internal network. Instead, “never trust, always verify” is the guiding tenet—even for machine-to-machine traffic within a facility. In practical terms, this means:
Identity-based access: Every user, device, and application must continuously authenticate and authorize access to network resources.
Least privilege enforcement: Access is always scoped to the minimum required for function, reducing the blast radius of potential compromise.
Continuous monitoring: Anomalous or unsanctioned behavior must be rapidly detected, reported, and acted upon.
ICS networks, however, bring unique barriers:
Legacy systems with minimal or no ability to run modern endpoint protections or identity agents.
High availability requirements where downtime for patching or reconfiguration is unacceptable.
Deterministic network requirements, where added latency or jitter may disrupt real-time process communications.
Microsegmentation Defined
Microsegmentation refers to the practice of creating finely-grained zones within the network, controlled at the application, workload, or even process level. This is fundamentally different from legacy segmentation at the IP subnet or VLAN level; microsegmentation enforces policy based on application identity, user/device attributes, and context.
Originally conceptualized in data center environments, practical microsegmentation is achieved through:
Software-defined networking (SDN)
Host-based firewalls and hypervisor-level controls
Identity-aware proxies and enforcement points
Persistent network flow analysis and policy enforcement
In ICS, these concepts must adapt to real-world constraints—especially the prevalence of devices lacking modern agents or APIs. Innovative approaches include protocol-aware gateways, inline enforcement appliances, and meticulous traffic baselining.
Applying Microsegmentation in ICS: Architecture and Best Practices
Dissecting the Purdue Model
Since the 1990s, the Purdue Model for ICS Security (Purdue Enterprise Reference Architecture) has provided a layered blueprint, separating enterprise, DMZ, and shop-floor assets. While helpful for initial segmentation, in practice, it is often too coarse. Within a given Purdue "zone" (e.g., Level 2—control devices, Level 1—field I/O), unrestricted lateral movement often persists.
Microsegmentation adds additional sub-zoning, for example:
Separating programmable logic controllers (PLCs) into logical groups with communication limited only to designated Human-Machine Interfaces (HMIs).
Restricting engineering workstations so they can only access authorized PLCs for maintenance windows, not continuously.
Implementing device-specific policies blocking unauthorized fieldbus communications.
Technical Building Blocks
Key components for ICS microsegmentation include:
Next-gen Firewalls with Deep Packet Inspection (DPI):
Modern security appliances parse industrial protocols (e.g., IEC-104, OPC UA) to enforce granular allow/deny rules—not just on IP/port, but on protocol function codes or values.
Identity-Based Segmentation:
Where possible, integrate device certificates, MAC-authentication, or network access control (NAC) to identify and segment endpoints even when IP addresses are dynamic or spoofable.
Enforcement Points at Multiple Layers:
Combine network-level controls (switch/router ACLs, firewalls) with in-line enforcement (protocol-aware proxies) and, where feasible, host-level firewalls for Windows/Linux SCADA and historian nodes.
Decoupling Operations:
Leverage zones of trust to enable maintenance and monitoring activities (e.g., remote engineering access) only through brokered, tightly-audited jump hosts.
Process Example: Securing a Water Treatment Plant ICS
In a typical water utility, you might deploy microsegmentation by:
Placing field controllers in their own subnet with access restricted solely to authorized SCADA HMIs.
Enforcing firewall rules that block all unnecessary services and protocols such as HTTP, or even unnecessary Modbus function codes.
Deploying an identity-aware proxy for vendor remote maintenance—forcing MFA and session recordings.
Implementing constant monitoring to alert on any anomalous communication attempts between previously isolated device groups.
IT/OT Collaboration: Overcoming Cultural and Technical Barriers
Microsegmentation’s efficacy depends not only on technical enforcement but also on the combined acumen of IT and OT personnel. Bridging these historically siloed domains is a recurrent challenge. Common issues include:
Misaligned Objectives: IT prioritizes confidentiality and flexible access, OT values uptime and process reliability.
Protocol and Asset Knowledge Gaps: IT staff may lack deep understanding of industrial protocols; OT teams may distrust outside interference with established processes.
Change Management: Traditional change control cycles can stall segmentation projects; cross-functional teams and rigorous testing can help mitigate friction.
Building trust through joint tabletop exercises, shared threat modeling, and involvement in microsegmentation pilots is paramount.
Deployment Considerations and Pitfalls
Asset Inventory and Dependency Mapping: Incomplete visibility leads to over-permissive rules or process disruptions. Tools like network flow analyses and passive discovery are core prerequisites.
Protocol Nuances: Industrial protocols lack robust security controls and may not respond well to DPI filtering. Rigorous vendor validation is mandatory before deploying protocol-aware security.
Legacy Devices: Some legacy controllers cannot even tolerate ARP changes or minor latency increases. Extensive lab testing—ideally with vendor cooperation—is required.
Policy Sprawl: Excessively granular policy without automation or rationalization quickly becomes unmanageable, leading to human error or catastrophic process impairment.
Conclusion: Microsegmentation as an Evolutionary Necessity
As targeted attacks transition from reconnaissance to full-scale disruption (see: Ukraine's 2015/2016 grid attacks, or Trisis/Triconex malware), no CISO or OT leader can afford the illusion of "trusted" internal networks. Microsegmentation is not a panacea—but it is a precondition for meaningful Zero Trust in ICS. Success depends as much on technical rigor as on operational empathy and cross-discipline collaboration.
ICS microsegmentation is, in essence, about limiting the blast radius—confining inevitably breached segments, slowing adversaries, and providing the breathing room needed for response and resilience.
Further Reading:
1. ISA/IEC 62443 Series: Security for Industrial Automation and Control Systems
2. NIST SP 800-82: Guide to Industrial Control Systems Security
3. MITRE ATT&CK for ICS
About the Author
Written by a security practitioner with 20+ years in the field of industrial cyber defense, having guided multiple energy sector segmentation deployments since the days of flat networks and crossed serial lines.