Employing bloom filters for privacy preserving distributed collaborative kNN classification

Increasingly, organizations are collecting users' personal data to mine rules that describe user behavior. In addition, different organizations may want to collaborate to derive rules based on collective data. However, due to the privacy-preserving requirements, organizations may not be able to share their data directly with others. In the current work, we employ Bloom filters to hide the sensitive data while still being able to perform collaborative data mining. In particular, we experiment with the kNN classifier. Using the Euclidean distance and Jaccard similarity measures, we evaluate the efficacy of the proposed representation. Using some real data, we show that Bloom filters effectively preserve data privacy while maintaining the accuracy of classifications.