Build an ISA/IEC 62443 Security Guard with OpenAI Agents SDK and Azure
Industrial control networks lack automated real-time compliance enforcement against unauthorized cross-zone Purdue model communications. Construct an OT threat detection and compliance agent using OpenAI Agents SDK deployed on Azure to monitor network traffic and trigger instant isolation playbooks.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an OT security monitoring agent leveraging OpenAI Agents SDK and Azure services to enforce ISA/IEC 62443 zone partitioning.
How work is evaluated
Tests agent ability to detect Purdue Level 0 to Level 3 violation packet sequences and issue compliant block rules.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
correct_clause_reference
Validates that the output references valid ISA/IEC 62443 clause.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
precision
Precision of illegal boundary violation flags • target: 0.99 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Parse Modbus and DNP3 packet logs across Purdue Model Level 1, 2, and 3 zones
Implement OpenAI Agents SDK runner with Azure AI search tools
Evaluate traffic against ISA/IEC 62443 security level (SL 1 to SL 4) guidelines
Automate target firewall block rule output for unauthorized Purdue boundary traversals
Reference links and supporting material
Network packet log samples containing standard SCADA protocols traversing Purdue network tiers.
Challenge setup
Tests agent ability to detect Purdue Level 0 to Level 3 violation packet sequences and issue compliant block rules.
Challenge input
JSON array of packet logs (src_ip, dest_ip, src_zone, dest_zone, protocol, payload)
OpenAI Agents SDK
Framework for modular multi-turn function calling agents
Azure
Cloud infrastructure for enterprise AI agent deployments
Evaluated output
JSON object with violation (boolean), iec_clause (string), dynamic_action (string)
- Validates that the output references valid ISA/IEC 62443 clause.
- Precision of illegal boundary violation flags • target: 0.99 • range: 0-1
- Benchmark: OT-Sec-IEC62443
- Precision target: 0.99
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- OpenAI · Required
- Azure · Optional
- Azure OpenAI Service · Optional
Evaluation contract
- correct_clause_reference · Weight 1
- precision · Weight 1
Metrics
- precision · Target 0.99
Benchmarks
- OT-Sec-IEC62443
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON array of packet logs (src_ip, dest_ip, src_zone, dest_zone, protocol, payload)
Output: JSON object with violation (boolean), iec_clause (string), dynamic_action (string)
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · agent_framework
- Azure · ai_engineering_tooling
Recipe freeze marker
No recipe freeze time is recorded.