Analysis and Proof-of-Concept Implementation of Private Computation Using an LPN-Based HFHE Library

This research examines the capabilities and limitations of the pvac_hfhe_cpp FHE library developed by Octra Labs through systematic source code analysis and the implementation of two Proof-of-Concepts (PoCs), namely private machine learning inference using polynomial regression and private information retrieval (PIR); the results confirm that the library functions correctly for arithmetic circuits while simultaneously exposing fundamental architectural constraints—specifically its symmetric-key and leveled design—which render it impractical for supporting complex, modern AI workloads, with the complete report and accompanying source code provided for reference.