IDCA NewsAll IDCA News
9 Mar 2022
Carnegie Mellon develops encryption for low-power IoT devices
Researchers at Carnegie Mellon University have developed new algorithms and hardware designs that enable low-power encrypted computing.
Smart devices such as smart phones, watches, sensors and other IoT devices generate an increasing amount of data that requires sophisticated computing, especially for extracting value from that data using machine learning. The devices could, instead, use computational offloading, sending sensor data to a nearby edge device or to the cloud for processing.
With offloading, even very sophisticated data processing is possible, but only with the caveat that the server performing the processing has unencrypted access to the data. Using a technology called homomorphically encrypted computing, these security concerns can be mitigated. The client encrypts its data, sends the encrypted data to the offloader, and the offloader processes it without ever decrypting the data.
It is expensive to offload encrypted workloads using traditional hardware designs, and only limited operations can be performed on encrypted data. Researchers at Carnegie Mellon developed new algorithms and hardware that directly address these costs, enabling encrypted offloading even on low-resource devices. As for the second drawback, the encrypted data can only be processed linearly, such as by adding and multiplying. The devices that use computational offloading usually send all of the data in one big package, and the servers carry out several computations at once. This consumes a lot of energy from the client. The researchers propose to send the encrypted data in smaller chunks, which would spread the energy demands over time.
Multiple rounds of communication with the server become possible as a result, according to the researchers. These new capabilities allowed the researchers to create processes that are most energy efficient for clients. A wide variety of IoT applications can take advantage of this new approach.
Photo credit: Christin Hume
Follow us on social media: