Privacy-Preserving Machine Learning as a Tool for Secure Personalized Information Services

Abstract

Abstract The article deals with the problems of cryptographic protection of data processing algorithms and techniques. They are novel techniques allowing to process private information without disclosing it to persons engaged in processing. One of the main applications of such security tools is the creation of personalized information services, which opens up new opportunities for business and reduces the risks of unauthorized access to personal data. We review important building blocks for cryptographic protection of data processing, such as zero-knowledge proofs, secure multi-party computations, and homomorphic encryption. Often, personalized information services are based on data mining and machine learning, so privacy-preserved machine learning is a very important building block for them. We analyze the concept of differential privacy which serves as the basis for privacy-preserving machine learning and some other cryptographic schemes. At the end of the paper, we forecast the perspectives of encrypted data processing.