Face Recognition and Adversarial Masking Techniques
Harsita Mav, Aneesh Mokashi, Srinath Nanduri, Vijaya U. Pinjarkar · 2022 3rd International Conference for Emerging Technology (INCET) · 2022
Face recognition is a system where people can rely on it to do various types of tasks from differentiating people to authentication. There are as many developments to it as the exploitations to it. The most common ones being physical ones which we can’t overcome, and adversarial attacks for which we can develop something to prevent them or remove them as a threat. The rise of deep learning and neural networks brought various opportunities and applications such as object detection and text-to-speech into modern society. Yet, despite the seemingly high accuracy, neural networks (and most machine learning models) could actually suffer from data, namely adversarial examples, that are manipulated very slightly from original training samples. In fact, past researches have indicated that as long as you recognize the "correct" method to vary your data, you'll force your network to perform poorly on data which may not seem to be visually. Our proposed system tries to overcome the problems faced by the existing models. It makes a face mesh for every person and stores it in a dataset remembering the faces, hence recognizing the faces in future using features from the mesh.