8 Privacy-Preserving Data Mining Removing identifying information from data doesnt work Even if the overtly identifying information can be removed, identification from remaining data is often possIble Data perturbation As discussed in Chapter 7, data perturbation can limit the privacy risks associated with the data without impacting analysis results Data mining often focuses on correlation and aggregation, both of which can generally be reliably accomplished with perturbed data rom Security in Computing, Fifth Edition, by Charles P. Pfleeger, et aL. (ISBN: 9780134085043) Copyright 2015 by Pearson Education, Inc. All rights reservedFrom Security in Computing, Fifth Edition, by Charles P. Pfleeger, et al. (ISBN: 9780134085043). Copyright 2015 by Pearson Education, Inc. All rights reserved. Privacy-Preserving Data Mining • Removing identifying information from data doesn’t work • Even if the overtly identifying information can be removed, identification from remaining data is often possible • Data perturbation • As discussed in Chapter 7, data perturbation can limit the privacy risks associated with the data without impacting analysis results • Data mining often focuses on correlation and aggregation, both of which can generally be reliably accomplished with perturbed data 8