Hacker News new | ask | show | jobs
by banterfoil 3114 days ago
I'm not choosing a side here, but saying "we have Wireshark" doesn't mean as much as you think it does. If packets are encrypted and sent at arbitrary times, we really have no idea what is being communicated. Surely we could see large amount of data if they were transferring entire audio sessions, but I feel like they are far more likely to send transcripts or key word mentions. Given the size of keyword mention, it would be very hard to distinguish this from a normal query to Google.
2 comments

The problem with that is processing power. If the Echo could do speech to text without the cloud (such as generating that transcript you mention), then it wouldn't need the cloud in the first place.

I'll give you that it could be possible to slip in other "trigger words" that could be tallied, then statistics sent off, but really, is that a problem? So what if your Echo counts how many times you say a specific keyword? Would that affect your life in any way?

Can't one just profile "Ok Google what is the temperature" versus the television being on? Or Music being played in the background?

it seems like it wouldn't be that hard to get some datapoints about what google might be sending based upon when transmissions were sent and what was being sent to the microphone.

Or just find someone smart enough to reverse engineer the device.