Secure and Efficient Outsourced k-Means Clustering using Fully Homomorphic Encryption with Ciphertext Packing Technique

In the era of cloud computing, the widespread adoption of machine learning applications, especially for sensitive tasks such as banking and healthcare, raises concerns about data privacy. Traditional cloud-based machine learning approaches expose sensitive data to potential misuse. This paper introduces an innovative solution leveraging Fully Homomorphic Encryption (FHE) for secure and efficient outsourced k-Means clustering. The proposed YASCHE algorithm employs FHE with a ciphertext packing technique, allowing parallel computation on encrypted data across multiple clouds. This significantly enhances the efficiency of the scheme, addressing the limitations of traditional Partial Homomorphic Encryption. By securing data through FHE, the system ensures privacy preservation during machine learning tasks, preventing unauthorized access and misuse of sensitive information.

Keywords:

Fully Homomorphic Encryption, Cloud Computing, Machine Learning, Privacy Preservation, k-Means Clustering, Outsourcing, Ciphertext Packing, Security, Data Privacy, Parallel Computation.