Synthetic Data Generation for Testing

[iAmGiG]

Goal

Generate synthetic IoT traffic data for controlled testing and validation.

Motivation

Why synthetic data?

1. Controlled experiments: Test specific attack patterns
2. Class balancing: Generate minority class samples
3. Edge case testing: Create rare but important scenarios
4. Privacy: Shareable data without privacy concerns
5. Research contribution: Synthetic IoT traffic generation is valuable

Approach

Phase 1: Tool Selection

Options (2025):

• SDV (Synthetic Data Vault): Best for tabular data
• CTGAN: GAN-based tabular synthesis
• SMOTE: Classic oversampling (simple, works well)

Recommendation: Start with SDV and CTGAN

Phase 2: Experiments

Experiment 1: Data Augmentation (synthetic + real)
Experiment 2: Generalization (train synthetic, test real)
Experiment 3: Edge Case Testing
Experiment 4: Privacy-preserving dataset publication

Deliverables

• Evaluate synthetic data tools (SDV, CTGAN, SMOTE)
• Implement data generation pipeline
• Validate synthetic data quality (statistical tests)
• Run augmentation experiments
• Run generalization experiments
• Document results
• Optional: Publish synthetic dataset

Timeline

6 weeks total (stretch goal)

Priority

STRETCH GOAL - High research value, not critical for core modernization