DPLog

An application of differential privacy to security monitorings

Intro

Privacy and security auditing are always seen as opposing forces, but does it have to be like that? Differential privacy offers the mathematical tools to strike a balance between privacy and auditing. This is why I have started to develop DPLog a host based agent that implements state of the art algorithms to allow security queries with privacy.

Components

The project is divided into the client which will run on multiple hosts and a server which will run on a single host.

Folder structure explained:

• client: the client component
• server: the server component written in Django
• docker: containerized version of the server only
• docs: documents
• screens: screenshots

Run server

The server currently runs on a file sqlite database.

Create a pyenv with python 3.6 and install the requirements:

• pip3 install -r requirements.txt

To run the server:

• remove the db.sqlite3
• manage.py check
• manage.py makemigrations
• manage.py migrate

Add an admin user with a password you want:

manage.py shell -c "from django.contrib.auth.models import User; User.objects.create_superuser('admin', 'admin@example.com', 'admin')"

Run the server:

• manage.py runserver

You should see something like this:

Django version 2.2, using settings 'controlstation.settings'
Starting development server at http://127.0.0.1:8000/
Quit the server with CTRL-BREAK.

Run client

Create a pyenv with python 3.6 and install the requirements:

• pip3 install -r requirements.txt

Compile the client to a windows binary by running:

compile.bat

Once the binary is compiled will be available in the dist folder.

You can then run it via the scheduler:

• start the task via scheduletask.ps1
• stop the task via stoptask.ps1