MIT’s new chip could stop hackers from stealing wearable tech data

Scientists from MIT have developed a new form of protection against a secret weapon of hackers the world over: the side-channel attack.

Their system is an integrated chip that defends against attacks while using much less energy than other comparable methods, a post from MIT News reveals.

A side-channel attack gathers private data by collecting indirect information such as power consumption from a system or its hardware. Worryingly, it could be used to gather health information from smartwatches and other data from unknowing users.

The MIT team’s new method is much less energy-intensive than other solutions available today, which is why it can be used for internet-of-things (IoT) devices such as smartwatches and other wearables. This chip is smaller than a thumbnail and can be placed into a smartphone, smartwatch, and other everyday electronic devices. Once integrated into a device, it performs secure machine-learning computation on sensor values, allowing it to detect any side-channel infiltrators.

Protecting user privacy with machine learning

The new system is a type of application-specific integrated circuit (ASIC) chip. The MIT researchers developed the system using a special type of computation known as threshold computing, meaning the data the chip reads is split into random components so that no side-channel data can be read by accessing the chip. This method is more computationally expensive, as the chip must now run more operations, but the researchers optimized the process so that it would require less multiplication to process the data.

“The goal of this project is to build an integrated circuit that does machine learning on the edge, so that it is still low-power but can protect against these side-channel attacks so we don’t lose the privacy of these models,” says Anantha Chandrakasan, the dean of the MIT School of Engineering, Vannevar Bush Professor of Electrical Engineering and Computer Science, and senior author of the paper.

“People have not paid much attention to security of these machine-learning algorithms, and this proposed hardware is effectively addressing this space.”

While the current implementation of their method does require 5.5 times more power and 1.6…