"Toward a Stronger Defense of Personal Data"

Researchers at the Massachusetts Institute of Technology (MIT) developed an Application-Specific Integrated Circuit (ASIC) chip to defend Internet of Things (IoT) devices against power-based side-channel attacks. A side-channel attack is a security exploit that seeks to gather information by measuring or exploiting the indirect effects of a system or its hardware rather than directly targeting a program or code. For example, in one type of side-channel attack, a skilled hacker could monitor a device's power consumption fluctuations while a neural network operates to extract protected information that leaks out of the device. Current methods developed to prevent some side-channel attacks are known to be power-intensive, and therefore, are not feasible for IoT devices such as smartwatches, which rely on lower-power computation. To address this issue, the MIT researchers built an integrated circuit chip capable of protecting IoT devices against power-based side-channel attacks while using significantly less energy than typical security techniques. Their chip, which is smaller than a thumbnail, could be implemented into a smartwatch, smartphone, or tablet to perform secure Machine Learning (ML) computations on sensor values. The researchers wanted to build an integrated circuit that performs ML on edge, so that it is low-power but still defensive against these side-channel attacks, and the privacy of these models is still preserved. The chip is based on a special type of computation called threshold computing, which involves splitting data into unique random components rather than having a neural network operate on the actual data. With the chip, the network operates on those different components individually in a random order before collecting the final result. Using this method, the information leakage from the device is random every time, thus preventing any actual side-channel information from being revealed. This article continues to discuss the new lower-energy chip developed by MIT researchers to prevent hackers from extracting hidden information from an IoT device. 

MIT reports "Toward a Stronger Defense of Personal Data"