Paper 2024/849

Fast, Large Scale Dimensionality Reduction Schemes Based on CKKS

Abstract

The proliferation of artificial intelligence and big data has resulted in a surge in data demand and increased data dimensionality. This escalation has consequently heightened the costs associated with storage and processing. Concurrently, the confidential nature of data collected by various institutions, which cannot be disclosed due to personal privacy concerns, has exacerbated the challenges associated with data analysis and machine learning model training. Therefore, designing a secure and efficient high-dimensional data reduction method that supports multi-party joint participation becomes critical to solving these problems. This paper proposes a novel homomorphic encryption dimensionality reduction scheme (HE-DR) based on CKKS, which modifies the Rank-Revealing (RR) method to make it more applicable to fully homomorphic encryption, thereby achieving fast and secure dimension reduction for high-dimensional data. Compared to traditional homomorphic encryption dimensionality reduction schemes, our approach does not transmit the user’s original data to other participants in any format (Ciphertext or Plaintext). Moreover, our method's computational efficiency is nearly $60-200$ times faster than similar algorithms, and the communication overhead is only $1/3$ of theirs. Finally, we have shown that our proposed scheme can preserve its computational efficiency and accuracy even when dealing with high-dimensional data. As dimensionality escalates, the ratio of ciphertext to plaintext computational efficiency plateaus at approximately 5 times, while the computational error (distance between subspaces) remains around $1e^{-11}$