It is not clear why we should necessarily be suspicious of smart TVs based on the findings here.
The lines seem to be blurred between what data is being sent by the TV OS itself versus a third party app that someone may have decided to use during the period that the researchers watched outbound network connections.
If someone is using the Netflix app, this necessitates that the TV must communicate with Netflix. How else would it work? Similarly, if you use the YouTube app, you would definitely get some requests to Google. Surely some developers of smart TV apps have decided to use Mixpanel or Adobe for analytics.
The "sensitive user data" alluded to in the headline is later admitted to be, at a minimum, "information about the device people are using, their locations, and possibly even when they are interacting with it".
Let's unpack that. It's pretty common for every web browser to send a user agent. Why shouldn't apps for a TV send device info to the app maker? Location can be inferred via IP, like any network request (nothing special here). And with any request, you of course know the time it was made (and could infer the interaction that produced it, like opening the app).
So they've told us nothing to support the headline. Indeed, the original paper from Northeastern notes that they did not try to inspect the data or MITM encrypted connections, so all we know is that requests are being made to these services during the course of operating the device and its apps. Not that the device manufacturer itself is sending your data, unprompted, to these third parties. So, the headline does not match the story.
Seems like university PR office must have gone a little crazy with this one. Why is nobody on HN questioning such a sensationalist, substanceless article?