<abstract><p>Private Set Intersection (PSI), which is a hot topic in recent years, has been extensively utilized in credit evaluation, medical system and so on. However, with the development of big data era, the existing traditional PSI cannot meet the application requirements in terms of performance and scalability. In this work, we proposed two secure and effective PSI (SE-PSI) protocols on scalable datasets by leveraging deterministic encryption and Bloom Filter. Specially, our first protocol focuses on high efficiency and is secure under a semi-honest server, while the second protocol achieves security on an economic-driven malicious server and hides the set/intersection size to the server. With experimental evaluation, our two protocols need only around 15 and 24 seconds respectively over one million-element datasets. Moreover, as a novelty, a <italic>multi-round</italic> mechanism is proposed for the two protocols to improve the efficiency. The implementation demonstrates that our <italic>two-round</italic> mechanism can enhance efficiency by almost twice than two basic protocols.</p></abstract>
Experimental evaluation of ensemble classifiers for imbalance in Big Data