CASE STUDY
Cutting Edge Identity Protection Using Machine Learning and Biometrics
Overview
TrustStamp is a computer vision and biometric startup that provides identity and trust as a Service to answer two fundamental questions: “Who are you?” and “Do I trust you?”. 10Clouds has been providing a staff augmentation service for TrustStamp for over 5 years, including Python developers specialized in Machine Learning and data processing.
10C engineers were highly involved in TrustStamp’s R&D projects. They worked with strong collaboration or under the supervision of the client's main Data Scientist. Still, developers had much independence in their day-to-day work, including both implementation and research.
Project Name
Trust Stamp
Website
Services
Biometric Authentication System, Machine Learning Algorithms, Design, Front-end Development
Type
Web Application
Industry
Our challenge
The most frequent problem was to find the encoding of an object. The encoding is a fixed-size vector of numbers. Some projects had also restrictions about the maximal size due to limited resources of the final device. Another class of problem was to detect malicious data and attacks on existing systems. Also, many smaller issues had to be solved while working on the whole algorithm like answering the questions: is the quality of a picture good enough? where is the object on an image? how to increase the amount of data and/or its quality?
How we made it happen
The solutions were usually deep neural networks trained and evaluated by our engineers. Especially convolutional neural networks (CNN). However, many classical Machine Learning algorithms were in use as well. We also involved methods from Computer Vision to pre- and post-process images but also to detect objects.