The act of signing/registration people up for participation is enrollment.
Since we have little control over devices such as cameras or sensors, the biometric template arrives as
plaintext. If we encrypt it immediately and only process it as ciphertext, we have the maximum practical
level of privacy. An important part of offering this highest level of privacy is a one-way encryption
algorithm, meaning that given ciphertext, there is no mechanism to get to the original plaintext. Many
one-way encryption algorithms exist, such as MD5 and SHA-512. However, these algorithms are not homomorphic.
This means we cannot do a closeness match between two ciphertext vectors using Euclidean measurements.
Open Inference offers a general purpose solution that produces biometric ciphertext that is Euclidean-measurable.
We do this using a Neural Network. We then apply a classification algorithm to allow for one-to-many
identification. This solution maximizes privacy and runs between O(1) and O(log(n)) time.
Enrollment is the act of introducing a Subject to the indentification or authenication system. Enrollment,
for Open Inference, is the process of extracting features and training a neural network for a particular
subject. For Enrollment to work reliably, we give as mnay biometric instances as possible. We use in
excess of 10 images, which the software morphs into 250 distinct images. The system then trains on 250
images. All of the training and images are encrypted BEFORE any form of training guaranteeing full privacy.