Physical-Cyber Anomaly Diagnostic System for Water Plants with AutoGen
Water treatment plants face silent physical threats like sensor tampering that occur without network security breaches. Implement a multi-agent conversational simulation using AutoGen where Physical Process Diagnostic and Network Security agents debate and isolate uncompromised physical attacks versus cyber intrusions.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an AutoGen multi-agent debate system to distinguish physical sensor spoofing from network cyber intrusions in municipal water networks.
How work is evaluated
Evaluates whether the AutoGen conversation accurately identifies physical sensor manipulation in scenarios with clean network logs.
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.
threat_classification_match
Verifies threat is correctly categorized as physical vs cyber.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
debate_convergence_turns
Number of conversation turns required to reach accurate consensus. • target: 4 • range: 1-10
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Build AutoGen multi-agent conversational loops between OT domain specialists
Simulate physical process dynamics vs network packet verification
Detect sensor drift and signal spoofing absent of IT/OT network compromise
Formulate localized safety isolation procedures for water plant operators
Reference links and supporting material
Paired network firewall logs and physical water flow/pH/pressure sensor streams from test water treatment plants.
Challenge setup
Evaluates whether the AutoGen conversation accurately identifies physical sensor manipulation in scenarios with clean network logs.
Challenge input
JSON containing network_log_status and physical_flow_delta_pct.
AutoGen
Enables multi-turn conversational debate and group consensus mechanisms.
Evaluated output
JSON containing threat_category, consensus_reached, and recommended_action.
- Verifies threat is correctly categorized as physical vs cyber.
- Number of conversation turns required to reach accurate consensus. • target: 4 • range: 1-10
- Benchmark: Cyber-Physical Threat Isolation Benchmark
- Debate Convergence Turns target: 4
- 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
- AutoGen · Required
- Microsoft AutoGen · Optional
Evaluation contract
- threat_classification_match · Weight 1
- debate_convergence_turns · Weight 1
Metrics
- debate_convergence_turns · Target 4
Benchmarks
- Cyber-Physical Threat Isolation Benchmark
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing network_log_status and physical_flow_delta_pct.
Output: JSON containing threat_category, consensus_reached, and recommended_action.
Execution harness: Python
Evaluation tools
- AutoGen · Multi-Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.