Computer Vision over Homomorphically Encrypted Data

Over the past decade, computer vision (CV) systems have become integral to healthcare, surveillance, and personal devices. The sensitive nature of data and models raises privacy concerns. Fully homomorphic encryption (FHE) allows computations on encrypted data, ensuring privacy. This tutorial explores integrating FHE into CV, addressing its challenges, mathematical foundations, key FHE schemes, SIMD capabilities, and hands-on demonstrations. It covers private and encrypted CV tasks and discusses open research directions.