A framework for efficient data anonymization under privacy and accuracy constraints
read more
Citations
No free lunch in data privacy
ACM Transactions on Database Systems
Protecting Sensitive Labels in Social Network Data Anonymization
ρ-uncertainty: inference-proof transaction anonymization
A Survey on Blockchain-Based Internet Service Architecture: Requirements, Challenges, Trends, and Future
References
k -anonymity: a model for protecting privacy
L-diversity: Privacy beyond k-anonymity
t-Closeness: Privacy Beyond k-Anonymity and l-Diversity
L-diversity: privacy beyond k-anonymity
Protecting respondents identities in microdata release
Related Papers (5)
Frequently Asked Questions (7)
Q2. What are the future works mentioned in the paper "A framework for efficient data anonymization under privacy and accuracy constraints" ?
In the future the authors plan to extend their framework to other privacy paradigms, such as t-closeness and m-invariance. Furthermore, the authors intend to study the privacy- and accuracy-constrained problems for data streams.
Q3. What is the metric used in the experiments?
In their experiments the authors use KL-Divergence (K LD), which has been acknowledged as a representative metric in the data anonymization literature [Kifer and Gehrke 2006].
Q4. What is the way to address the inflexibility of single-dimensional recoding?
To address the inflexibility of single-dimensional recoding, Mondrian [LeFevre et al. 2006a] employs multidimensional global recoding, which achieves finer granularity.
Q5. What is the common way to achieve k-anonymity?
k-anonymity is commonly achieved either by generalization (e.g., show only the area code instead of the exact phone number) or suppression (i.e., hide some values of the quasi-identifier), both of which inevitably lead to information loss.
Q6. Why is the data in Mondrian not close to each other?
Because Mondrian uses space partitioning, the data points within a group are not necessarily close to each other in the QT space (e.g., points 22 and 55 in Figure 1(b)), causing high information loss.
Q7. What is the NCP of class G over all quasiidentifier attributes?
The NCP of class G over all quasiidentifier attributes isNCP(G) = d∑i=1 wi · NCPAi (G), (1)where d is the number of attributes in QT (i.e., the dimensionality).