F1-AI [πve]
Privacy-preserving Formula 1 race simulator using Fully Homomorphic Encryption (FHE) and threshold cryptography to compute car velocities on encrypted data without revealing car characteristics.
Demo Video
Project Information
At a Glance
Privacy-preserving Formula 1 race simulator using Fully Homomorphic Encryption (FHE) and threshold cryptography to compute car velocities on encrypted data without revealing car characteristics.
Description
F1-AI is an innovative racing simulation system that leverages advanced cryptographic techniques to maintain complete privacy during computation. The system uses Fully Homomorphic Encryption (FHE) to enable Formula 1 race simulations where car characteristics and performance metrics remain encrypted throughout the entire computation process. Five independent judges collaboratively generate encrypted car data through distributed key generation, ensuring no single party can access sensitive information. The system computes car velocities using the quadratic form S = t^T W t on encrypted data, then uses threshold decryption requiring all judges to reveal final race results. Cars can be "trained" through encrypted parameter updates to improve performance, all while maintaining cryptographic privacy guarantees.
Technical Details
The system implements the BFV (Brakerski-Fan-Vercauteren) fully homomorphic encryption scheme via OpenFHE, combined with threshold cryptography for distributed trust. Each car is represented by a 10-dimensional encrypted characteristic vector t and a 10×10 performance matrix W, generated collaboratively by 5 judges holding distributed key shares. Velocity computation evaluates the quadratic form S = Σᵢ Σⱼ tᵢ · Wᵢⱼ · tⱼ homomorphically on encrypted ciphertexts, producing Enc(S) without ever decrypting intermediate values. The threshold decryption protocol requires all 5 judges to participate, providing security under an honest-but-curious adversary model where collusion of fewer than all judges reveals no private information. Training modifies encrypted car parameters through homomorphic addition of encrypted deltas, creating improved car versions while preserving privacy.
Team
5Gabriele Lozupone
Emanuele Nardone
Emanuele Vita
Sara Marchetti
Rosa Fera