SECRETA is a system for evaluating and comparing anonymization algorithms for relational, transaction, and R-T (relational-transaction) datasets. It integrates 9 popular anonymization algorithms for relational and transaction datasets as well as 3 bounding methods for combining these algorithms, under a common, benchmark-oriented framework. This framework enables data publishers to apply and thoroughly analyze the performance of these algorithms. The results of the analysis are presented within an interactive visualization framework. More specifically, SECRETA supports Incognito, Cluster, Top-down, and Full subtree bottom-up algorithms for datasets with relational attributes, and COAT, PCTA, Apriori, LRA and VPA algorithms for datasets with transaction attributes. Additionally, it supports the RMERGEr, TMERGEr, and RTMERGEr bounding methods, which enable the anonymization of RT-datasets by combining two algorithms, each designed for a different attribute type (e.g., Incognito for relational attributes and COAT for transaction attributes).
SECRETA operates in two modes, namely Evaluation and Comparison.
The Evaluation mode can be used to configure and evaluate the effectiveness of a given algorithm, with respect to data utility and privacy, as well as its efficiency. For capturing data utility, SECRETA employs several information loss measures and supports data utility requirements. These requirements can be expressed using queries and/or utility constraints, which are specified by data publishers or generated automatically. Furthermore, SECRETA enables the use of 20 different combinations of relational and transaction anonymization algorithms to anonymize RT-datasets. Each combination of algorithms can be applied by using one of the three available bounding methods, thereby leading to a total of 60 different configurations. The selection and management of these configurations is performed in an intuitive way that allows preserving different aspects of data utility.
The Comparison mode offers data publishers the ability to design and execute benchmarks for comparing multiple anonymization algorithms. These benchmarks facilitate an interactive and progressive comparison of sets of algorithms, with respect to their utility and efficiency. The results of the comparative analysis are summarized and presented graphically, allowing for fast and intuitive understanding of the effectiveness and efficiency of different algorithms.
The architecture of the SECRETA system (shown in the figure) comprises a set of components which we broadly divide into frontend and backend components. The frontend offers a Graphical User Interface (GUI), which enables users to: (a) issue anonymization requests, and (b) visualize and store experimental results. The backend consists of components for servicing anonymization requests and for conducting experimental evaluations. The frontend of SECRETA is implemented using the QT framework, and the screens were designed using QT Creator . All the plots were created using the QWT library. The backend of our system is implemented in C++.