FHEID：情報漏洩ゼロの生体テンプレート、暗号化埋め込みベクトル検索、高信頼プライベート証明

FHEIDは、生体認証向け完全準同型暗号（FHE）フレームワークを開発し、暗号化された512次元特徴量ベクトルに対するユークリッド距離およびコサイン類似度照合を、生体テンプレートの漏洩なしにクラウド上で実行します。

FHEID Homomorphic Biometric Pipeline & Cloud Matching Topology

Client biometric extraction, CKKS homomorphic encryption, and blind cloud similarity search

Central System: FHE Biometric Engine - Encrypted Vector Matching Core

Cryptographic Nodes

  • Client Feature Extractor
  • CKKS Homomorphic Encrypter
  • Untrusted Cloud Matcher
  • Encrypted Threshold Evaluator
  • Client Decision Decryptor

Architectural Layers & Protocols

  • Client Biometric Extraction Plane (Edge Neural Network / Secure Enclave): Extracting normalized 512-element floating-point feature embeddings from iris, face, or fingerprint sensors.
  • Homomorphic Vector Cryptosystem (CKKS Ring-LWE Homomorphic Encryption): Encrypting biometric embeddings into high-degree polynomial rings with configurable noise budgets.
  • Blind Cloud Evaluation Engine (SIMD Vector Batching & Rescaling Operations): Computing encrypted dot products across millions of stored encrypted templates in parallel without decryption.
  • Privacy Compliance & Audit Layer (GDPR Article 9 / CCPA / BIPA Biometric Shield): Guaranteeing mathematical impossibility of biometric template reconstruction from stolen databases.

Key Metrics: Zero Decryption (Cloud Biometric Privacy) | < 85 ms (Homomorphic Match Time) | 512 Dimensions (Vector Embedding Size) | 100% (GDPR / BIPA Compliance)

プライバシーを保護するFHE生体認証、信頼されないクラウドエンクレーブでの暗号化虹彩・指紋ベクトルマッチング

Authoritative technical engineering reference for privacy-preserving fhe biometric authentication, encrypted iris/fingerprint vector matching in untrusted cloud enclaves in NIST FIPS 203/204/205 standards, lattice-based cryptography, side-channel attack mitigation, and quantum-resilient infrastructure.

Core Engineering Areas

Technical Articles