Embedded Software Implementation of Privacy Preserving Matrix Computation using Elliptic Curve Cryptography for IoT Applications

Security concerns about data collected and processed by network-connected electronic devices has motivated the need for privacy-preserving computation. One of the most promising theoretical tools in this direction is homomorphic encryption. In this work, we explore the efficient implementation of elliptic curve-based additively homomorphic encryption which allows certain matrix-vector operations in the encrypted domain without the prohibitive computational complexity of fully homomorphic encryption. We present an algorithmic framework for the same, supported by detailed performance analysis of our software implementation evaluated on an embedded microprocessor along with memory-time trade-offs and side-channel countermeasures. We also discuss various applications which can benefit from this privacy-preserving compute framework.