Excuse me for reposting:
We are conducting a study where we need to keep audio recordings of participant evaluations. Can anyone suggest how the voices in these recordings could be distorted so they could be anonymised? I thought it is an easy problem, but similar questions on several platforms remain unanswered. #privacy #anonymity @dataGovernance @datasovereignty
@vanderbeeken @dataGovernance @datasovereignty hm, I'm not so sure. Certainly the type of data we collect is not so sensitive to justify including resourceful actors in a possible threat model of ours.
As for transcripts, if they were readily available and not a sufficient proxy for the recording, converting them in audio wouldn't solve the problem
@mapto @dataGovernance @datasovereignty Thank you. It is hard to comment, as I don’t know anything about the sensitivity of your data, or about what anynomization you would like to achieve.
Here is a recent article:
https://www.eff.org/deeplinks/2023/11/debunking-myth-anonymous-data
@vanderbeeken @dataGovernance @datasovereignty it is I that should thank you for considering my issue. In this other response I quoted one of the studies mentioned in your article. https://qoto.org/@mapto/111407685657723169
The scope of my intended anonymisation is not to publicly distribute the recordings and give unlimited time to any potential attacker to try to deanonymise them, but rather, in case at any point we need to refer back to the original data, we can show a particular point of interest as evidence of our data collection for auditing purposes.
@mapto @dataGovernance @datasovereignty Is that even possible in the age of AI? Safest bet: have someone else read the transcripts.