On the privacy preserving properties of random data perturbation techniques

By Hillol Kargupta, Souptik Datta, Qi Wang, Krishnamoorthy Sivakumar. Third IEEE International Conference on Data Mining, 2003

791 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Mathematical Physics, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/baef0d749f6b2537bccc0cc380641a50e1ac6081

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Mathematical Physics

In Statistics and Probability

Citations in the corpus

A Survey of Quantification of Privacy Preserving Data Mining Algorithms (2008, background): “According to [16], additive random noise can be filtered out u sing certain signal processing techniques with very high accuracy.”
The applicability of the perturbation based privacy preserving data mining for real-world data (2008, methodology): “Since then many good approaches have been proposed [3] [5] [6].”
Future Data and Security Engineering (2014, methodology): “com corpora [11] (with 1,600,000 tweets) show that our forests are more accurate than baseline algorithms, including Multinomial Naı̈ve Bayes (MNB [12]), Maximum Entropy (MaxEnt [13]) and SVM [10].”
Toward Privacy in Public Databases (2005, background): “[24] argue that independent additive perturbation may have limited application to preserving privacy insofar as certain informative features of the data set (e.”
PDA: Privacy-Preserving Data Aggregation in Wireless Sensor Networks (2007, background): “Privacy preservation has also been studied in the data mining domain [18], [19], [20], [21].”
Mining Association Rules under Privacy Constraints (2008, background): “This approach was extended in [3] and [26] to address a variety of subtle privacy loopholes.”
Privacy-Preserving SVM Classification on Vertically Partitioned Data (2006, background): “However, there is some debate about the security properties of such algorithms [9, 10].”
Worst-Case Background Knowledge for Privacy-Preserving Data Publishing (2007, background): “For example, spectral techniques can separate much of the noise from the data if the noise is uncorrelated with the data [17, 19].”
Privacy preserving data classification with rotation perturbation (2005, methodology): “…inference); (2) approximately reconstructing the data from the perturbed data and then estimating the original data from the reconstructed data [8] [6] (approximation-based inference); and (3) if the distributions of the original columns are known, the values or the properties of the values…”
L-diversity: Privacy beyond k-anonymity (2007, neutral): “On the negative side, [47] shows that randomizing the data, especially by adding zero mean random variables, does not necessarily preserve privacy.”
A Comprehensive Survey of Privacy-preserving Federated Learning (2021, methodology): “As a disadvantage, the data perturbation may degrade the data utility and may be vulnerable to a noise reduction [86].”
Privacy-preserving SVM using nonlinear kernels on horizontally partitioned data (2006, background): “However, there is some debate about the security properties of such algorithms [17].”
A Random Rotation Perturbation Approach to Privacy Preserving Data Classification (2005, background): “The class label attribute of the data tuple is public, i.e., privacy-insensitive.”
Privacy-preserving Naïve Bayes classification (2008, background): “pointed out an important issue: arbitrary randomization is not safe [25].”
L-diversity: privacy beyond k-anonymity (2006, neutral): “On the negative side, [ Kargupta et al. 2003 ] shows that randomizing the data, especially by adding zero mean random variables, does not necessarily preserve privacy.”