kp-anonymity

take a look at the presentation!

Paper

This work is based on the novel approach for the anonymization of time series with a special focus on the pattern loss, presented by this paper

Datasets

• Sales transactions over a year - UCI
• News social feedback over eight months - UCI

Requirements

foo@bar:~$ python3 -m pip install -r requirements.txt

• numpy==1.18.1
• pandas==1.0.3
• loguru==0.5.1
• saxpy==1.0.1.dev167 (https://github.com/seninp/saxpy)
• pathlib==1.0.1
• matplotlib==3.2.2

How to launch the tool

foo@bar:~$ python3 kp-anonymity.py algorithm k_value p_value paa_value dataset_path dataset_output_path

Explain Parameters

• kp-anonimity : main program
• algorithm : choose between naive or kapra approach
• k_value : value of k-anonymity
• p_value : value of p-anonymity, pattern
• paa_value : to reduce the dimensionality of patterns (see how)
• dataset_path : csv input file
• dataset_output_path : csv output file

Example

foo@bar:~$ python3 kp-anonymity.py kapra 10 2 5 Dataset/Input/Sales_Transaction_Dataset_Weekly_Final.csv Dataset/Anonymized/output.csv

Time Utility

To compare time scalability between naive and kapra approaches, launch the test utility:

foo@bar:~$ cd Utility
foo@bar:~$ ./test.sh

Repository Structure

• Dataset: contains the datasets used in my tests
  • Input: input datasets for the tool
  • Anonymized: store the output of the tool
• Paper: contains the two papers studied for this project
  • Utility-Based Anonymization for Privacy Preservation with Less Information Loss
  • Supporting Pattern-Preserving Anonymization for Time-Series Data
• Utility: contains scripts for verify time efficiency and for resetting the tool
• kp-anonymity.py: main script, which implement naive and kapra algorithms
• node.py: manage the create-tree phase of both algorithms
• dataset_anonymized.py: manage the printing and anonymized value replacing of the output dataset
• create_dataset.py: script for generating subdataset of the News Social Dataset, used for measure time scalability of both algorithms
• requirements.txt: list of necessary packages

Author and Credits

Author: Giorgio Rossi, student of Computer Engineering (LM) - UNIGE - a.y. 2019/2020.

Final project of the course Data Protection and Privacy. Work based on Davide Caputo's repository.