Machine Learning and Privacy at the Edge with Edge Impulse’s Daniel Situnayake

2022-11-23

Download 422

Edge devices are hardware devices that sit at the edge of a network. They could be routers, switches, your phone, voice assistant, or even a sensor in a factory that monitors factory conditions. Machine learning on the edge combines ideas from machine learning with embedded engineering. With machine learning models running on edge devices amazing new types of applications can be built, such as using image recognition to only take pictures of the objects you care about, developing self-driving cars, or automatically detect...

Edge devices are hardware devices that sit at the edge of a network. They could be routers, switches, your phone, voice assistant, or even a sensor in a factory that monitors factory conditions.

Machine learning on the edge combines ideas from machine learning with embedded engineering. With machine learning models running on edge devices amazing new types of applications can be built, such as using image recognition to only take pictures of the objects you care about, developing self-driving cars, or automatically detect potential equipment failure.

However, with more and more edge devices being used all the time that might be collecting sensitive information via sensors, there are a number of potential privacy and security concerns.

Dan Situnayake, Head of Machine Learning at Edge Impulse, joins the show to share his knowledge about the practical privacy and security concerns when working with edge IoT devices and how to still leverage this incredible technology but do so in an ethical and privacy-preserving way.

Topics:

What’s your background and how did you end up as the head of machine learning at Edge Impulse?
What is an edge device?
What is Edge Impulse and what are the types of use cases people are solving with AI on edge devices through the Edge Impulse platform?
What are the unique security challenges with edge devices?
Since these devices are potentially observing people, collecting information about someone’s movements, what kind of privacy concerns does someone building for these devices need to think about?
Are there industry best practices for protecting potentially sensitive information gathered from such devices?
Is there research into how to collect data but protect someone's privacy when it comes to building training sets in machine learning?
What happens if someone steals one of these devices? Are there safeguards in place to protect the data collected on the device?
Where do you see this industry going in the next 5-10 years?
Do you foresee security and privacy getting easier or harder as these devices become more and more common?

Resources: