Mobile AI applications increasingly rely on cloud servers to classify images, but sending raw images to the cloud creates serious privacy concerns. A new paper introduces PRICE, a framework designed to address this problem in mobile edge-cloud computing environments.
PRICE leverages homomorphic encryption, a technique that permits computation directly on encrypted data. This means the cloud could potentially perform image classification without ever seeing the original pixels, keeping user data private while still enabling AI-powered services.
The paper's abstract highlights the promise of homomorphic encryption for this purpose, though the full details of the PRICE framework—including its architecture and any trade-offs—are presented in the full manuscript. The work points toward a future where privacy-sensitive mobile AI applications can use cloud resources without compromising user trust.