Privas

Privas - Assuring privacy in data exploring systems

Privas key points:

A deep analysis of the current state-of-the-art in data privacy and an onthology representing this research. This analysis allowed to have a better overview of the evolution of the data privacy preservation and the different ways to achieve this preservation, and to build an onthologyto represent all these terms and relationships;
The documentation of the data exploration process with identification of multiple rolesin this process as well as their concerns and responsibilities;
A language to describe the repository to preserve the privacy, the information containedin it, and the level privacy preservation. This PrivasL language was designed after having analyzed the information required toapply the preservation;
A modular and scalable framework that allows to anonymize different kinds of repositorieswith different privacy preservation levels.This modular framework was built using Java and ANTLR. Each module can growindependently as is decoupled by design;
A decision engine to choose what type of Privacy Model apply depending on the conditionsbeing specified;

The technology of nowadays allows to easily extract, store, process and use information about individuals and organizations. The increase of the amount of data collected and its value to our society was, at first, a great advance that could be used to optimize processes, find solutions and support decisions but also brought new problems related with lack of privacy and malicious attacks to confidential information.

Privas enables, among other things, to anonymize databases. It should be used by data publishers to protect information from attacks controlling the desired privacy level and the data usefulness. In order to specify these requirements a DSL (PrivasL) is used and the automatization of repository transformation, that is based on language processing techniques, is the novelty of this work.

The tool

The Privas tool was built as a two separate pieces of software.

As the main component, Privas framework that enables to preserve the privacy in different types of repositories, by using a description of the repository, and the privacy needs (PrivasL description).

The web platform offers a visual tool to describe the repositories and its needs and to generate the PrivasL description.

To better understand each one of these components the Dissertation file that produced this work is advised.

Next, for each component, it is given the steps to compile and run them.

The framework

Compile the framework

                mvn package

The privas framework can be compiled using the maven package compilation.

Run framework

                java -jar target/privas.jar FILENAME_PRIVASL

The privas framework should be started using the java compiler command, using the parameters of the PrivasL file.

Example

PrivasL description

                CSV
    Path: "fileCSV.csv"

file
        data [ &age, &work_class, &education_num, &marital_status,
            &occupation, &race, &sex, &native_country ]

Prevent from: [ recordLinkage ]

To run the example you can start by downloading this census CSV.

The PrivasL description of the CSV can be found on the right. This information can also be downloaded here.

To use this description, please change its contents so the path of the CSV is the same as saved in your computer. For example if you have saved it in: /users/me/Downloads/census.csv Change the PrivasL description content with: Path: "/users/me/Downloads/census.csv"

Run privas with census CSV

                java -jar target/privas.jar /users/me/Downloads/census.csv

This example, can then be executed with the command on the right.

Privas, with this specification will transform the CSV and anonymize the age information.

The web platform

The Privas' web platform was designed as a separate application.

Its purpose is to generate PrivasL files through a visual framework.

This platform can also load existing PrivasL and perform changes to it.

Compilation

Compile the web platform

                ./gradlew build

The webplatform should be compiled using the maven package command. This will generate the executable needed to run the platform

Running

Run the web platform

                ./gradlew bootRun

The platform should be then runned using the java command presented on the right.

To access the web platform, it should be available in:

http://localhost:8080

Screenshots

Below there are some screenshots of the web platform.

Ontology

During the research, a deep analysis of the current state-of-the-art in data privacy was performed. This analysis allowed to have a better overview of the evolution of the data privacy preservation and the different ways to achieve this preservation. In the end, and as a way to group this knowledge, multiple ontologies were built. These ontologies represent the different concepts and relationships existing in data privacy.

Next a simplified version of a full ontology is present. In the end there also exist links to consult the full ontology and the two other different ontologies produced.

A simplified PPDP ontology

Other ontologies

A general ontology of PPDP

An ontology of Data

An overall privacy ontology

Other information

The Privas work:

Privas - Automatic Anonymization of databases
Joana Margarida Miguel, Maria João Varanda Pereira, Pedro Rangel Henriques and Mario Berón, International Conference on Applied Computing 2019, pp. 29-38, 2019, ISBN: 978-989-8533-95-1
Considered among the best papers of International Conference on Applied Computing 2019 conference and was selected to be published -- in a extended version -- in the IADIS International Journal on Computer Science and Information Systems
Assuring data privacy with Privas - A tool for data publishers
Joana Margarida Miguel, Maria João Varanda Pereira, Pedro Rangel Henriques and Mario Berón, IADIS International Journal on Computer Science and Information Systems, Vol. 14, No. 2, pp. 41-58, ISSN: 1646-3692

Privas was developed during a master thesis.

It was built as a modular software, to allow future extensions. The ways Privas can be extended is by:

Adding more types of repositories (Currently MySQL and CSV are available)
Adding more types of privacy models (Currently k-anonymity, l-diversity, t-closened and e-differential privacy are supported)
Adding different techniques to the current/new available privacy models
Enhance the decision engine to apply the privacy techniques