Open jcomeauictx opened 7 years ago
in response to my objection that the security holes will likely be patched soon:
Since we only need vibration / pattern information below 10 Hz, patches to avoid speech recognition will not affect our ability to automatically start the App.
In terms of the identification by fingerprinting, it will be difficult to fix this with an update. Web browser fingerprinting has proven to be hard to avoid and most people who try to do it give up due to the inconvenience.
In general, users don’t seem to be willing to make much of an effort to protect privacy. It will also be easy to get the permissions needed, even if the firmware is patched.
So, it is likely that this implicit data will be available in phones for years, at least on many phones unless some standard is imposed. Finally, if users are asked if they want the app to start automatically when it is needed, most will agree and this could then set the needed permissions , especially if no privacy issue is raised by the way we use the data.
At our meeting, I mentioned that when we release Myturn for general use, it would be desirable to have it start automatically when it was needed. One approach would be to detect a conversation pattern indicating a need for facilitation. We know from previous research that unbalanced switching pauses indicate a lack of warmth in a conversation. (switching pause = the time between when one person stops talking and another person starts talking) According to this recent publication, unbalances like this would be easy to detect:
https://crypto.stanford.edu/gyrophone/
This approach would make it unnecessary to use the phone’s microphone, thus protecting privacy.
The publication also mentions a method of identifying phones implicitly, thereby permitting the type of reputation development needed once we go beyond a one-button system.
Having these capabilities implicitly makes the transition from the niche application we will be testing to a mass market application a lot easier.
dss