edyan / neuralyzer

Neuralyzer is a library and a command line tool to anonymize databases (by updating existing data or populating a table with fake data)

edyan/neuralyzer

Summary

This project is a library and a command line tool that anonymizes a database by updating data or generating fake data (update vs insert). It uses Faker to generate data from rules defined in a configuration file.

As it can do row per row or use batch mechanisms, you can load tables with dozens of millions of fake records.

It uses Doctrine DBAL to abstract interactions with databases. It's then supposed to be able to work with any database type. Currently it works (tested extensively) with MySQL, PostgreSQL and SQLServer.

Neuralyzer has an option to clean tables by injecting a DELETE FROM with a WHERE critera before launching the anonymization (see the config parameters delete and delete_where).

Neuralyzer had an option to clean tables but it's now managed by pre and post actions :

entities:
    books:
        cols:
            title: { method: sentence, params: [8], unique: true }
        action: update
        pre_actions:
            - db.query("DELETE FROM books")
post_actions:
    - db.query("DELETE FROM books WHERE title LIKE '%war%'")

Installation as a library

composer require edyan/neuralyzer

Installation as an executable

You can even download the executable directly (example with v3.1):

$ wget https://github.com/edyan/neuralyzer/raw/v4.0/neuralyzer.phar
$ sudo mv neuralyzer.phar /usr/local/bin/neuralyzer
$ sudo chmod +x /usr/local/bin/neuralyzer
$ neuralyzer

Usage

The easiest way to use that tool is to start with the command line tool. After cloning the project and running a composer install, try:

$ bin/neuralyzer

Generate the configuration automatically

Neuralyzer is able to read a database and generate the configuration for you. The command config:generate accepts the following options:

Options:
    -D, --driver=DRIVER              Driver (check Doctrine documentation to have the list) [default: "pdo_mysql"]
    -H, --host=HOST                  Host [default: "127.0.0.1"]
    -d, --db=DB                      Database Name
    -u, --user=USER                  User Name [default: "www-data"]
    -p, --password=PASSWORD          Password (or it'll be prompted)
    -f, --file=FILE                  File [default: "neuralyzer.yml"]
        --protect                    Protect IDs and other fields
        --ignore-table=IGNORE-TABLE  Table to ignore. Can be repeated (multiple values allowed)
        --ignore-field=IGNORE-FIELD  Field to ignore. Regexp in the form "table.field". Can be repeated (multiple values allowed)

Example

bin/neuralyzer config:generate --db test_db -u root -p root --ignore-table config --ignore-field ".*\.id.*"

That produces a file which looks like:

entities:
    authors:
        cols:
            first_name: { method: firstName, unique: false }
            last_name: { method: lastName, unique: false }
        action: update # Will update existing data, "insert" would create new data
        pre_actions: {  }
        post_actions: {  }

    books:
        cols:
            name: { method: sentence, params: [8] }
            date_modified: { method: date, params: ['Y-m-d H:i:s', now] }
        action: update
        pre_actions: {  }
        post_actions: {  }

guesser: Edyan\Neuralyzer\Guesser
guesser_version: '3.0'
language: en_US

You have to modify the file to change its configuration. For example, if you need to remove data while anonymizing and change the language (see Faker's doc for available languages), do :

# be careful that some languages have only a few methods.
# Example : https://github.com/FakerPHP/Faker/tree/v1.14.1/src/Faker/Provider/fr_FR
language: fr_FR

INFO: You can also use delete in standalone, without anonymizing anything. That will delete everything in books:

entities:
    authors:
        cols:
            first_name: { method: firstName, unique: false }
            last_name: { method: lastName, unique: false }
        action: update
    books:
        pre_actions:
            - db.query("DELETE FROM books")

If you wanted to delete everything then insert 1000 new books:

guesser_version: '3.0'
entities:
    authors:
        cols:
            first_name: { method: firstName, unique: false }
            last_name: { method: lastName, unique: false }
        action: update
    books:
        cols:
            name: { method: sentence, params: [8] }
        action: insert
        pre_actions:
            - db.query("DELETE FROM books")
        limit: 1000

Run the anonymizer

To run the anonymizer, the command is simply "run" and expects:

Options:
    -D, --driver=DRIVER      Driver (check Doctrine documentation to have the list) [default: "pdo_mysql"]
    -H, --host=HOST          Host [default: "127.0.0.1"]
    -d, --db=DB              Database Name
    -u, --user=USER          User Name [default: "www-data"]
    -p, --password=PASSWORD  Password (or prompted)
    -c, --config=CONFIG      Configuration File [default: "neuralyzer.yml"]
    -t, --table=TABLE        Do a single table
        --pretend            Don't run the queries
    -s, --sql                Display the SQL

    -m, --mode=MODE          Set the mode : batch or queries [default: "batch"]

Example

bin/neuralyzer run --db test_db -u root -p root

That produces that kind of output:

Anonymizing authors
 2/2 [============================] 100%

Queries:
UPDATE authors SET first_name = 'Don', last_name = 'Wisoky' WHERE id = '1'
UPDATE authors SET first_name = 'Sasha', last_name = 'Denesik' WHERE id = '2'

....

WARNING: On a huge table, --sql will produce a HUGE output. Use it for debugging purpose.

Library

The library is made to be integrated with any Tool such as a CLI tool. It contains:

• A Configuration Reader and a Configuration Writer
• A Guesser
• A DB Anonymizer

Guesser

The guesser is the central piece of the config generator. It guesses, according to the field name or field type what type of faker method to apply.

It can be extended very easily as it has to be injected to the Writer.

Configuration Writer

The writer is helpful to generate a yaml file that contains all tables and fields from a DB. A basic usage could be the following:

<?php

require_once 'vendor/autoload.php';

// Create a container
$container = Edyan\Neuralyzer\ContainerFactory::createContainer();
// Configure DB Utils, required
$dbUtils = $container->get('Edyan\Neuralyzer\Utils\DBUtils');
// See Doctrine DBAL configuration :
// https://www.doctrine-project.org/projects/doctrine-dbal/en/2.7/reference/configuration.html
$dbUtils->configure([
    'driver' => 'pdo_mysql',
    'host' => '127.0.0.1',
    'dbname' => 'test_db',
    'user' => 'root',
    'password' => 'root',
]);

$writer = new \Edyan\Neuralyzer\Configuration\Writer;
$data = $writer->generateConfFromDB($dbUtils, new \Edyan\Neuralyzer\Guesser);
$writer->save($data, 'neuralyzer.yml');

If you need, you can protect some cols (with regexp) or tables:

<?php
// ...
$writer = new \Edyan\Neuralyzer\Configuration\Writer;
$writer->protectCols(true); // will protect primary keys
// define cols to protect (must be prefixed with the table name)
$writer->setProtectedCols([
    '.*\.id',
    '.*\..*_id',
    '.*\.date_modified',
    '.*\.date_entered',
    '.*\.date_created',
    '.*\.deleted',
]);
// Define tables to ignore, also with regexp
$writer->setIgnoredTables([
    'acl_.*',
    'config',
    'email_cache',
]);
// Write the configuration
$data = $writer->generateConfFromDB($dbUtils, new \Edyan\Neuralyzer\Guesser);
$writer->save($data, 'neuralyzer.yml');

Configuration Reader

The configuration Reader is the exact opposite of the Writer. Its main job is to validate that the configuration of the yaml file is correct then to provide methods to access its parameters. Example:

<?php
require_once 'vendor/autoload.php';

// will throw an exception if it's not valid
$reader = new Edyan\Neuralyzer\Configuration\Reader('neuralyzer.yml');
$tables = $reader->getEntities();

DB Anonymizer

The only anonymizer currently available is the DB one. It expects a PDO and a Configuration Reader objects:

<?php

require_once 'vendor/autoload.php';

// Create a container
$container = Edyan\Neuralyzer\ContainerFactory::createContainer();
$expression = $container->get('Edyan\Neuralyzer\Utils\Expression');
// Configure DB Utils, required
$dbUtils = $container->get('Edyan\Neuralyzer\Utils\DBUtils');
// See Doctrine DBAL configuration :
// https://www.doctrine-project.org/projects/doctrine-dbal/en/2.7/reference/configuration.html
$dbUtils->configure([
    'driver' => 'pdo_mysql',
    'host' => '127.0.0.1',
    'dbname' => 'test_db',
    'user' => 'root',
    'password' => 'root',
]);

$db = new \Edyan\Neuralyzer\Anonymizer\DB($expression, $dbUtils);
$db->setConfiguration(
    new \Edyan\Neuralyzer\Configuration\Reader('neuralyzer.yml')
);

Once initialized, the method that anonymize the table is the following:

<?php
public function processEntity(string $entity, callable $callback = null): array;

Parameters:

• Entity: such as table name (required)
• Callback (callable / optional) to use a progress bar for example

A few options can be set by calling :

<?php
// Limit of fake generated records for updates and creates.
// Default : 0 = everything to update / nothing to insert
public function setLimit(int $limit);
// Don't do anything, default true
public function setPretend(bool $pretend);
// Return or not a result, default false
public function setReturnRes(bool $returnRes);

Full Example:

<?php

require_once 'vendor/autoload.php';

// Create a container
$container = Edyan\Neuralyzer\ContainerFactory::createContainer();
$expression = $container->get('Edyan\Neuralyzer\Utils\Expression');
// Configure DB Utils, required
$dbUtils = $container->get('Edyan\Neuralyzer\Utils\DBUtils');
// See Doctrine DBAL configuration :
// https://www.doctrine-project.org/projects/doctrine-dbal/en/2.7/reference/configuration.html
$dbUtils->configure([
    'driver' => 'pdo_mysql',
    'host' => 'mysql',
    'dbname' => 'test_db',
    'user' => 'root',
    'password' => 'root',
]);

$reader = new \Edyan\Neuralyzer\Configuration\Reader('neuralyzer.yml');

$db = new \Edyan\Neuralyzer\Anonymizer\DB($expression, $dbUtils);
$db->setConfiguration($reader);
$db->setPretend(false);
// Get tables
$tables = $reader->getEntities();
foreach ($tables as $table) {
    $total = $dbUtils->countResults($table);

    if ($total === 0) {
        fwrite(STDOUT, "$table is empty" . PHP_EOL);
        continue;
    }
    fwrite(STDOUT, "$table anonymized" . PHP_EOL);

    $db->processEntity($table);
}

Pre and Post Actions

You can set an array of pre_actions and post_actions that will be executed before and after neuralyzer starts to anonymize an entity.

These actions are actually symfony expressions (see Symfony Expression Language) that rely on Services. These Services are loaded from the Service/ directory.

For now there is only one service : Database that contains a method query usable like that : db.query("DELETE FROM table").

Configuration Reference

bin/neuralyzer config:example provides a default configuration with all parameters explained :

config:

    # Set the guesser class
    guesser:              Edyan\Neuralyzer\Guesser

    # Set the version of the guesser the conf has been written with
    guesser_version:      '3.0'

    # Faker's language, make sure all your methods have a translation
    language:             en_US

    # List all entities, theirs cols and actions
    entities:             # Required, Example: people

        # Prototype
        -

            # Either "update" or "insert" data
            action:               update

            # Should we delete data with what is defined in "delete_where" ?
            delete:               ~ # Deprecated (delete and delete_where have been deprecated. Use now pre and post_actions)

            # Condition applied in a WHERE if delete is set to "true"
            delete_where:         ~ # Deprecated (delete and delete_where have been deprecated. Use now pre and post_actions), Example: '1 = 1'
            cols:

                # Examples:
                first_name:
                    method:              firstName
                last_name:
                    method:              lastName

                # Prototype
                -

                    # Faker method to use, see doc : https://fakerphp.github.io/
                    method:               ~ # Required

                    # Set this option to true to generate unique values for that field (see faker->unique() generator)
                    unique:               false

                    # Faker's parameters, see Faker's doc
                    params:               []

            # Limit the number of written records (update or insert). 100 by default for insert
            limit:                0

            # The list of expressions language actions to executed before neuralyzing. Be careful that "pretend" has no effect here.
            pre_actions:          []

            # The list of expressions language actions to executed after neuralyzing. Be careful that "pretend" has no effect here.
            post_actions:         []

Custom application logic

When using custom doctrine types doctrine will produce an error that the type is not know. This can be solved by providing a bootstrap file to register the custom doctrine type.

bootstrap.php

<?php

require_once '../vendor/autoload.php';

\Doctrine\DBAL\Types\Type::addType('custom_type', 'Namespace\Of\The\Custom\Type');

Then provide the bootstrap file to the run command:

bin/neuralyzer run --db test_db -u root -p root -b bootstrap.php

Development

Neuralyzer uses Robo to run its tests (via Docker) and build its phar.

Clone the project, run composer install then...

Run the tests

• Change the --wait option if you have a lot of errors because DB is not ready.
• Change the --php option for 7.2 or 7.4
• Set --no-coverage if you want to disable PHPUnit Code Coverage.

With MySQL

$ vendor/bin/robo test --php 7.2 --wait 10 --db mysql --db-version 5
$ vendor/bin/robo test --php 7.3 --wait 10 --db mysql --db-version 8
$ vendor/bin/robo test --php 7.4 --wait 10 --db mysql --db-version 8
$ vendor/bin/robo test --php 8.0 --wait 10 --db mysql --db-version 8

With PostgreSQL 9, 10 and 11 (12 also works)

$ vendor/bin/robo test --php 7.2 --wait 10 --db pgsql --db-version 10
$ vendor/bin/robo test --php 7.3 --wait 10 --db pgsql --db-version 11
$ vendor/bin/robo test --php 7.4 --wait 10 --db pgsql --db-version 12
$ vendor/bin/robo test --php 8.0 --wait 10 --db pgsql --db-version 13

With SQL Server

Warning : 2 tests fail, because of strange behaviors of SQL Server ... or Doctrine / Dbal. PHPUnit can't compare 2 Datasets because the fields are not in the same order.

$ vendor/bin/robo test --php 7.2 --wait 15 --db sqlsrv
$ vendor/bin/robo test --php 7.3 --wait 15 --db sqlsrv
$ vendor/bin/robo test --php 7.4 --wait 15 --db sqlsrv
$ vendor/bin/robo test --php 8.0 --wait 15 --db sqlsrv

Build a release (with a phar and a git tag)

$ php -d phar.readonly=0 vendor/bin/robo release

Build the phar only

$ php -d phar.readonly=0 vendor/bin/robo phar

Improve code quality with phpinsights

docker run -it --rm -v $(pwd):/app nunomaduro/phpinsights analyse --fix

Update dependencies to make sure it'll work with PHP 7.2

vendor/bin/robo composer:update