AnonTool

Introduction

In database systems one of the possible solutions to protect our data is to anonymize it; but performing a proper anonymization is a really complex process. The majority of existing solutions uses the common techniques to "sanitize" the data: detecting identifiers and quasi-identifiers, properly aggregating and/or removing them. These solutions do not guarantee complete anonymization, so they cannot be the best choise. Recently, a new mechanism to anonymize data, the Differential Privacy, has becoming the dominant way to perform data anonymization, because it provides a quantified privacy level with a rigorous mathematical foundation. Our tool is based on this technique. With this tool we target to simplify the technical details and provide an easy-to-use and easy-to-deploy system to anonymously release your data.

Description

This project implements a Web service to provide anonymized queries on statistical data. In particular, when deployed, through HTTP requests, it allows to upload files on which users can make queries. It provides a Web interface to make use of Google C++ library differential-privacy.

System requirements:

• Architecture: 64-bit host or virtual machine
• OS: Linux

File requirements:

• file extension must be .csv,
• each entry must have a unique ID (multiple occurences of the same ID are not allowed),
• there must be at least a numeric column on which it is possible to compute queries,
• numeric values must be in range [0, 150].

An example can be found in differential-privacy-master/differential-privacy/operations/data.csv.

Query requirements:

• only SELECT queries are supported with single statistical operations. Allowed operations are the followings:
  • COUNT,
  • SUM,
  • AVG,
  • VAR,
  • STDEV,
  • MAX,
  • MIN,
• only operations on a single column are permitted and the choosen column must contain only numbers,
• queries cannot contain WHERE, HAVING, ORDER BY clauses.

Some examples can be found in queries_examples.json.

Working features:

• If you want to change the IP addresses or ports, you have to modify the file docker-compose.yml.
• In the file Const.py, you can find all the system parameters, such as HTTP routes, etc. [WARNING: Be careful when changing differential privacy parameters]

Build

Follow these steps to launch docker containers for simulate application:

1. Install docker
2. Install docker-compose
3. Open terminal and place in project folder
   1. (OPTIONAL) In order to change server IP address and port in the container, modify diff_priv_server fields in docker-compose.yml
   2. in DiffPrivServer.py defaults are set to 172.25.0.2:5002
4. To create containers with docker-compose use command:
   • docker-compose up -d --build (ONLY FOR FIRST RUN)
   • docker-compose up -d --build --force-recreate -t 0
   • [NOTE] If docker raises some errors:
     • retry command using sudo
     • type on terminal sudo service docker start and retry command
5. Open the browser and choose to go to:
   • 172.25.0.2:5002 to see all the available HTTP routes
   • 172.25.0.2:5002/index to use the tool

6. To shutdown containers with docker-compose use command:
   • docker-compose down

Possible errors

1. ERROR: for diffprivtool_diff_priv_server_1 Cannot start service diff_priv_server: driver failed programming external connectivity on endpoint diffprivtool_diff_priv_server_1 (...): Error starting userland proxy: listen tcp 0.0.0.0:5002: bind: address already in use

   Solution:

   • on host command line, type lsof -i :5002
   • if some process is using the port, type kill [PID] to kill it

2. EXISTING FILES

   • In order to remove existing files (.csv, logs, users_list) from the host machine, delete diffpriv_server folder in /var/lib/docker/volumes/