This post is the transcript of an informational video I created. You can find the video at the bottom of the post.
In the last few years we are seeing an explosion of devices to track our health and fitness.
Two important reasons for the explosion are:
Sensors are becoming cheaper and cheaper, and
Nearly everyone now walks around with an internet gateway in their pocket
These two facts together mean that companies can cheaply create devices that record data from sensors on the body and ship that data to their servers.
There, algorithms can learn from new data, and process it to return useful statistics to the user, like calories burned, blood pressure, or even cortisol levels. This is currently the leading paradigm in fitness and health wearable technology.
Now, if the last decade of internet has taught us something, is that entrusting sensitive personal data to the cloud is a guaranteed recipe for disaster. To give one eminent wearable-related example: in February 2018, hackers stole data from 150 million users of the popular fitness app “MyFitnessPal”. This app tracks diet and exercise, and it can be connected to many wearable fitness devices, collecting the data they generate.
If your instinctive reaction to third parties having access to your health metrics is “so what?”, you’re probably not exercising your imagination enough. Data gets breached all the time for all kind of service providers. Combine the breaches together, and things get uncomfortable very fast.
What happens when a third party can combine, for example, your DNA data with your diet and fitness metrics, credit data, and political affiliation? (This is not speculation, huge leaks of all these types of data have appeared in the news in the last few years). What happens is that they can start predicting things about you, that you yourself are unaware of. Obviously not a desirable state of things, especially considering the kind of players that would buy data from illegal hacks.
What’s the alternative?
We must migrate to service architectures where sensitive data is not centralized, but rather kept with the user. This challenge can be tackled by combining Edge computing and Privacy-preserving Machine Learning.
With Edge computing, we process data locally rather than on the cloud. In this way, user privacy is protected, and there’s no central repository of sensitive data there to tempt hackers and shady players.
To keep improving the quality of the service, we can use recent developments in Privacy-preserving Machine Learning algorithms to learn from a great number of users, while they only provide data which can’t be used to reconstruct sensitive information.
We’ll dive in more details on Privacy-preserving Machine Learning in a future episode of this series.