Project Highlights:
- Microservice-based architecture and WebRTC
- Machine learning for biometric recognition
- Development team to contribute product ideas
Face & Voice Recognition and Authentication Solution
Client and Business Goals:
A US-based company tasked us with the development of an enterprise verification-as-a-service (EVaaS) solution for securing access to sensitive data.
Product Description:
A microservice-based single sign-on (SSO) software that enables biometric authentication using face and voice recognition.
Applied Technologies:
Python 3, Django, Django REST Framework, Flask, Celery, MoviePy, ffmpeg, OAuth2, Java, Amazon S3, PostgreSQL, Redis, CI/CD (Docker, Docker Swarm, Jenkins), JavaScript, React, WebRTC (OpenTok), BarCode readers, UI/UX design, manual and automated API testing
Data science and Machine Learning: TensorFlow, Keras, dlib, OpenCV, Google Vision.
Microservice-Based Architecture and WebRTC
- Each microservice becomes a separate subproject with its own functionality, which makes it easier to write, support, and enhance. This also allowed us to select optimal solutions and frameworks according to the goals and specifics of each microservice.
- The biometric authentication flow includes facial capture and voice recognition. An off-the-shelf WebRTC service was implemented as there was a need to process media data to the server.
Machine Learning for Biometric Recognition
- The product has to be able to identify the user based on voice, photo, and questions.
- As per usual, we were desperately in need of valid data. We collected initial data sets of voice and photo entries. We have also evaluated over 10 available solutions to validate US driver’s licenses and choose an OCR module.
- It was important that we could see the accuracy of these solutions provided in real-world scenarios. We elected to go with Google Vision as we saw it as the best option for the product to integrate with.
Development Team to Contribute Product Ideas
- Despite its enterprise origin, it was crucial to bring a startup spirit to the development of this product. We owe our success to the fact that our team was immersed in the project and focused on thinking of big ideas and finding legitimate solutions. Features like face anti-spoofing, question generation using NLP or extra voice, and photo data sets all came from the team and eventually become part of the product.
Deep Learning
For questions generation
Web RTC
For real-time data
Face Anti-Spoofing
Techniques
The final product became highly customizable, with all the internal complexity hidden under a simple user interface. Our client's partnering businesses are currently able to integrate this solution with their own products via API. Another feature worth noting is single sign-on, in order to spread single authentication across multiple enterprise systems for the convenience of users.
