Fogger - GDPR friendly database masker

Purpose

Fogger is a tool that solves the problem of data privacy. When developers need to work with production data but are obliged to comply with GDPR regulations they need a way to get the database copy with all the sensitive data masked. And while you can always write your own, custom solution to the problem - you don't have to anymore - with fogger you are covered.

Apart from masking data you can also subset or even exclude some tables. Don't worry for the relations with foreign keys, fogger will refine database so everything is clean and shiny.

You can configure various masking and subsetting strategies, and when what fogger has to offer is not enough - you can easily extend it with your own strategies.

How to use the docker image

Fogger requires docker environment, redis for caching and two databases: source and target. You can set up this stack using for example this docker-compose file:

version: '2.0'
services:
  fogger:
    image: tshio/fogger:latest
    volumes:
    - .:/fogger
    environment:
      SOURCE_DATABASE_URL: mysql://user:pass@source:3306/source
      TARGET_DATABASE_URL: mysql://user:pass@target:3306/target
      REDIS_URL: redis://redis
  worker:
    image: tshio/fogger:latest
    environment:
      SOURCE_DATABASE_URL: mysql://user:pass@source:3306/source
      TARGET_DATABASE_URL: mysql://user:pass@target:3306/target
      REDIS_URL: redis://redis
    restart: always
    command: fogger:consumer --messages=200
  redis:
    image: redis:4
  source:
    volumes:
    - ./dump.sql:/docker-entrypoint-initdb.d/dump.sql
    environment:
      MYSQL_DATABASE: source
      MYSQL_PASSWORD: pass
      MYSQL_ROOT_PASSWORD: pass
      MYSQL_USER: user
    image: mysql:5.7
  target:
    environment:
      MYSQL_DATABASE: target
      MYSQL_PASSWORD: pass
      MYSQL_ROOT_PASSWORD: pass
      MYSQL_USER: user
    image: mysql:5.7

Note:

we are mapping volume to fogger's and worker's /fogger directory - so the config file would be accessible both in container and in our host filesystem
we are importing database content from dump.sql

Of course you can modify and adjust the settings to your needs - for example - instead of importing database from dump file you can pass the existing database url to fogger and worker containers in the env variables.

Now we can spin up the set-up by docker-compose up -d. If the database is huge and you want to speed up the process you can spawn additional workers executing docker-compose up -d --scale=worker=4 instead. Give it a few seconds for the services to spin up then you can start with Fogger:

Fogger gives you three CLI commands:

docker-compose run --rm fogger fogger:init will connect to your source database and prepare a boilerplate configuration file with the information on tables and columns in your database. This configuration file is a place where you define which column should be masked (and how) and which tables should be subsetted. See [example config file](Example config file).
docker-compose run --rm fogger fogger:run is the core command that will orchestrate the copying, masking and subsetting of data. The actual copying will be done by background worker that can scale horizontally. Before run is executed, make sure that the config file has been modified to your needs. Available subset and mask strategies has been described below.
docker-compose run --rm fogger fogger:finish will recreate indexes, refine database so that all the foreign key constraints are still valid, and then recreate them as well. This command runs automatically after run so you need to execute it only when you have stopped the run command with ctrl-c.
it's done - the masked and subsetted data are in a target database. You can do whatever you please with it. For example: docker-compose exec target /usr/bin/mysqldump -u user --password=pass target > target.sql will save the dump of masked database in your filesystem.

Example config file

tables:
  posts:
    columns:
      title: { maskStrategy: starify, options: { length: 12 } }
      body: { maskStrategy: faker, options: { method: "sentences" } }
    subsetStrategy: tail
    subsetOptions: { length: 1000 }
  comments:
    columns:
      comment: { maskStrategy: faker, options: { method: "sentences" } }
  users:
    columns:
      email: { maskStrategy: faker, options: { method: "safeEmail" } }
excludes:
    - logs

This is an example of config file. The boilerplate based on your database schema will be generated for you by fogger:init, all you have to do is fill in the mask strategies on the columns that you want masked and subset strategies on the tables for which you only want fraction of the rows.

For the clarity and readability of the config files, all the tables that will not be changed can be omitted. They will be copied as they are. Similarly you can omit columns that are not to be masked. Tables from the excludes section will exist in the target database, but will be empty.

List of available strategies

Masking data

hashify - will save the MD5 hash instead of data - you can pass optional argument: template

email: { maskStrategy: "hashify", options: { template: "%s@example.com" } }
starify - will save the 10 stars instead of data - you can pass optional argument: length to override default 10

email: { maskStrategy: "starify", options: { }
faker - will use a marvelous faker library. Pass the method of faker that you want to use here as an option.

email: { maskStrategy: "faker", options: { method: "safeEmail" } date: { maskStrategy: "faker", options: { method: "date", arguments: ["Y::m::d", "2017-12-31 23:59:59"] }

Subsetting data

range - only copy those rows, where column is between min and max

subsetStrategy: range
subsetOptions: { column: "createdAt", min: "2018-01-01 00:00", max: "2018-01-31 23:59:59" }

head and tail - only copy length first / last rows

subsetStrategy: head
subsetOptions: { length: 1000 }

subsetStrategy: tail
subsetOptions: { length: 1000 }

Under the hood

If you are interested what really happens:

source database schema without indices and foreign keys is copied to target
data is divided into chunks (this includes query modification for subsetting). Chunks are processed by background workers (using RabbitMQ)
during copying sensitive data is substituted for masked version - in order to keep the substituted values consistent, redis is used as a cache
when all data is copied, fogger will recreate indices
refining cleans up database removing (or setting to null) relations that point to excluded or subsetted table rows
the last step is to recreate foreign keys

Contributing

Feel free to contribute to this project! Just fork the code, make any updates and let us know!

TheSoftwareHouse/fogger