In reality, the neural network analysing the health information uses private data that might be utilised by a malevolent agent through a side-channel attack.
A side-channel attack is a method of obtaining secret information by utilising a system or its associated hardware. In fact, one sort of side-channel attack uses a neural network to extract protected data from the device’s power consumption oscillations. Movies show people trying to access locked safes.
The lock clicks as it is turned, indicating that rotating it in the same way will assist unlock the safe. This is a side-channel assault. It includes using unwanted data to forecast device behaviour.
However, existing solutions to avoid side-channel attacks take a lot of power, making them unsuitable for IoT devices like smartwatches.
A research team developed an integrated circuit device that can withstand power side-channel assaults while using far less energy than other security techniques. Incorporated inside a smartphone, smart watch, or tablet, the chip enables secure machine learning computations on sensor values.
The initiative’s goal is to build a low-power, side-channel-protected integrated circuit that executes machine learning at the edge.