Homomorphic Encryption for VR Motion Telemetry

This repository demonstrates a small CKKS homomorphic-encryption pipeline for privacy-preserving analytics over VR motion telemetry.

The current scripts focus on one real-valued telemetry feature, controllerSpeed, from normalized user-motion CSV files. The goal is to test whether a telemetry field can be encrypted and prepared for privacy-preserving aggregate computation without exposing raw motion values during the analytics step.

Why this exists

XR systems can produce sensitive behavioral traces: controller speed, hand motion, timing, reach patterns, and repeated interaction behavior. Even simple telemetry can become identifying when collected at scale.

This project explores a narrow research question:

Can real-valued VR telemetry be prepared for encrypted analytics using CKKS-style approximate homomorphic encryption?

Project overview

testscript.py   Local SEAL/Python wrapper smoke test using a sample normalized CSV
he.py           Encrypts controllerSpeed values across normalized user/session files
requirements.txt

Expected input layout:

chunk1/
├── user1/
│   └── session1_normalized.csv
├── user2/
│   └── session2_normalized.csv
└── ...

Each CSV is expected to contain:

controllerSpeed

What is encrypted

The controllerSpeed column is encrypted with CKKS parameters suitable for demo-scale approximate arithmetic:

• Polynomial modulus degree: 8192
• Coefficient modulus bits: [60, 40, 40, 60]
• Scale: 2^40

Encrypted values are kept in memory as ciphertext objects. The current demo does not persist ciphertext batches to disk.

Tech stack

• Python 3.9+
• Microsoft SEAL via Python bindings / wrapper
• CKKS approximate homomorphic encryption
• pandas for CSV loading
• pybind11 for native binding support

How to run

Install Python dependencies:

pip install -r requirements.txt

Ensure Microsoft SEAL and its Python bindings are installed and available in the active environment.

Run a local setup smoke test:

python testscript.py

Run the encryption pipeline:

python he.py

Case study

See docs/case-study.md for the research context, implementation decisions, and current limitations.

Current limitations

• This is a research prototype, not a production privacy system.
• The pipeline currently focuses on one real-valued feature.
• Ciphertexts are not persisted or passed into a complete encrypted aggregation service.
• CKKS introduces approximate numerical behavior; results require careful interpretation.
• Production use would require security review, parameter validation, key management, and threat modeling.

Future work

• Add a small synthetic sample dataset so the pipeline can be tested without private telemetry.
• Add encrypted aggregate examples such as mean controller speed.
• Persist ciphertext outputs in a documented format.
• Add a diagram connecting this repository to the broader VR telemetry privacy workflow.
• Add unit tests around CSV validation and parameter setup.

License

MIT License — use and adapt with attribution.