Foresight: Real Time Facial Detection and Recognition Using WebAssembly and Localized Deep Neural Networks

Abstract

The emergence of facial recognition technology is an appealing solution to address the many present-day needs for verification of identity claims. The application of such technology has become publicly available and has proved its effectiveness as an added layer of security through native applications. Until now, there has been no previous attempt at bringing this solution to a web-based platform supporting real time classification. With frequent reports of websites being exploited and databases being leaked, there is an urgent need for an intelligent security mechanism to overlay the current traditional authentication methods, including usernames and passwords. This paper investigates the possibility of unobtrusive, continuous authentication for web applications based on facial data collected using WebAssembly driven detection algorithms. This novel detection technique has proved the viability to perform real-time image processing on the web with the possibility of achieving near native speeds. This is accompanied with a competitive server side facial recognition rate of 91.67% achieved on the Labeled Faces in the Wild dataset.