Privacy-Preserving ID3 Algorithms: A Comparison

Many real-life scenarios require the analysis of large amounts of data from multiple sources. Often, the data contain highly sensitive information and may be subject to privacy laws preventing its aggregation and sharing. Privacy-preserving data mining has emerged as a solution to this problem. It enables data scientists to analyze the distributed data without having to place it in a central location and while guaranteeing its privacy. Decision tree classification is a popular and widely studied machine learning technique for which many privacy-preserving versions exist. In this paper, we review recent privacy preserving implementations of the ID3 classification technique in a distributed environment and compare them in terms of efficiency and privacy. We consider cases where data is split horizontally over multiple parties.