Our technology applies special masking algorithms to create a unique abstract form of the data that:
Keeps the exact inference model accuracy
KPIs of Confusion matrix, Accuracy, Percision & Recall and F1 score for a clear text data model and a masked data model are 99.99% equal
Does not require additional compute resources
Comprehensive enough to be processed by cloud-based Machine Learning and other data analytics tools services without paying the compute power penalty involved in processing encrypted data.
Additional computation time involved for processing a dataset of ~1GB is negligible (< 10 seconds).