Kumu Knowledge Graph JSON Generator

🎯 Goal

This Python script reads structured CSV files containing cybersecurity techniques, preconditions, and postconditions, then generates a Kumu-compatible JSON blueprint to build a semantic knowledge graph in the Kumu visualization tool.

The graph includes:

• Nodes for Techniques, Preconditions, and Postconditions
• Edges showing:
  • Preconditions belonging to Techniques
  • Techniques resulting in Postconditions
  • Semantic relationships between Postconditions and linked Preconditions

---

📂 Required CSV Input Files

The script requires three specifically formatted CSV files placed in the same directory:

1. Formatted relationships.csv

Defines the core techniques with associated preconditions and postconditions.

Tech ID	Tech Name	List of Precond	List of PostCond
D3-CIA	SomeTech	Precondition A; Precondition B	Postcondition X; Postcondition Y

2. Precondition.csv

Contains a list of preconditions with IDs and descriptions.

index	precondition
D3-CIA-I1	System must be isolated

3. Postcondition.csv

Contains postconditions with linked precondition references.

index	Postcondition	Matching Preconditions from ChatGPT
D3-CIA-C1	Network traffic is encrypted	D3-CIA-I1, D3-CIA-I2

💡 Note:

• Column headers must exactly match the above.
• The index column in Precondition.csv and Postcondition.csv is renamed during processing for clarity.

---

⚙️ How the Code Works

Step-by-Step Breakdown:

1. Read and Rename Columns
   Loads and normalizes column names across the three CSV files for consistency.

2. Map Data
   Creates internal mappings:

   • Technique_ID ↔ Technique_Name
   • Precondition_ID ↔ Precondition_Description
   • Postcondition_ID ↔ Postcondition_Description

3. Construct Relationships

   • Links preconditions to techniques
   • Links techniques to postconditions
   • Links postconditions to preconditions (semantic matching)

4. Generate Kumu Elements and Connections
   Builds a list of elements and connections using:

   • add_element() for node creation
   • add_connection() for edge creation with deduplication

5. Export JSON
   The final structure is saved as:

   {
     "elements": [...],
     "connections": [...]
   }
   

---

📄 Output File Format

The script outputs a file named: kumu_graph_complete.json

Example Structure:

{
  "elements": [
    {
      "label": "D3-CIA(SomeTech)",
      "type": "Technique",
      "description": "SomeTech"
    },
    {
      "label": "D3-CIA-I1",
      "type": "Postcondition",
      "description": "System must be isolated"
    },
    {
      "label": "D3-CIA-C1P",
      "type": "Precondition",
      "description": "Network traffic is encrypted"
    }
  ],
  "connections": [
    {
      "from": "D3-CIA(SomeTech)",
      "to": "D3-CIA-I1",
      "type": "is_postcondition_for",
      "direction": "directed"
    },
    {
      "from": "D3-CIA-C1P",
      "to": "D3-CIA(SomeTech)",
      "type": "results_in_postcondition",
      "direction": "directed"
    },
    {
      "from": "D3-CIA-I1",
      "to": "D3-CIA-C1P",
      "type": "semantically_links_to",
      "direction": "directed"
    }
  ]
}

---

🧴 Usage

Run the script from a terminal or notebook: python your_script_name.py

If all files are correctly formatted and available, it will output the JSON file: kumu_graph_complete.json You can now import this file into Kumu under the "Data" → "Import JSON Blueprint" section.

---

📝 Notes

• All labels are validated to avoid duplication or invalid entries.
• Missing condition IDs or mismatches are logged with warnings but skipped to prevent failure.
• Elements are automatically deduplicated.

---

🛠️ Dependencies

Ensure you have the following installed: pip install pandas If using in a Jupyter environment, IPython.display is also required for DataFrame preview.