Sentry

Introduction

This tool was built using deep learning, a KVM network to host all of the services, and a front-end integration mechanism.

Identity on the internet lacks a standard greater than a trustless third-party token or password. There is nothing more fundamental than biometrics to authenticate a user. Mouse tracking cannot be replicated by another person, as a users hand movements are intricately tied to a persons identity.

Using deep learning, we can extract a users mouse tracking behavior across an XY axis, reconstruct those plotted points as an image, and train a convolutional neural network to determine if new mouse movement patterns are authentic or not.

Having been hand engineered and built from the ground up, this tool was intended to be integrated into large enterprise applications where authentication is still somewhat archaic.

System Overview

The system used real-time websockets to track user mouse metrics. The entire system was hosted on a KVM, on-premises network, internalized to the enterprise using it without accessing external services.

Image Reconstruction