Make Answers Anonymous in the Statistics Tab

[brianrodri]

We'll accomplish this using k-anonymity. From Wikipedia:

A release of data is said to have the k-anonymity property if the information for each person contained in the release cannot be distinguished from at least k - 1 individuals whose information also appear in the release.

We'd like the statistics tab to have this property to protect learners from being identified. Anonymity can give our learners confidence in playing our explorations, and would improve the overall privacy of Oppia.

As far as implementation goes, we should only show a particular answer in the stats tab when k unique learners have input that same answer. Additionally, we should only show such answers if there are at least j of them in total, so something like:

def anonymize_answers(answers):
    anon_answers = [a for a in answers if a.unique_learners >= K]
    return anon_answers if len(anon_answers) >= J else []

The value of k and j should be kept in the admin config panel, and should thus be customizable (and testable) through there.

[ctao5660]

Hey @brianrodri , I might have some free time to take this up. Would this be a change to stats_services or a different file?

[brianrodri]

@ctao5660 that'd be great, thanks!

Yes, there are backend apis there where we should be able to filter out answers from.

[ctao5660]

@brianrodri I'm creating a ConfigProperty in StatsServices for the k and j values so it can be changed through the admin config panel, what do you think are good k and j values to start with?

[brianrodri]

Let's start with k = 5 and j = 3. Thanks!

[ctao5660]

@brianrodri there already seems to be a constant regulating the k value called feconf.STATE_ANSWER_STATS_MIN_FREQUENCY should the k value override this? Or should we replace the constant in feconf with the correct k value?

[brianrodri]

@ctao5660 Hmm, that value isn't sufficient for k, because it's an aggregate of all learner answers (a single learner could put in k answers). We need to base k on the number of unique learners who have input an answer (maybe through unique exploration attempts? I'm not sure if we have a way to identify unique learners yet)

[ctao5660]

@brianrodri Hello, I am going to be away for the next week, I can deassign myself from the issue if this is something that needs to be urgently worked on.

[brianrodri]

No problem, it is high priority so I'll deassign just in case someone else wants to make an attempt while you're away. Feel free to reassign yourself once you're available again!

[anamsarfraz]

I am interested in working on this issue for GHC OSD

[iamprayush]

Hi @brianrodri! Is this free to work on?