https://bugzilla.wikimedia.org/show_bug.cgi?id=58196

--- Comment #41 from Tisza GergÅ‘ <[email protected]> ---
> To avoid strange statistical variance like that, I'd recommend making 
> timestamp of the currently running month appear as last month's.

I don't see how that would help (other than making the window of opportunity
smaller). You could still poll the data on the first day of the month to get an
accurate edit timestamp. Since the attack is based on reading *when* the
timestamp changes and correlating that with edit times, *what* the timestamp
changes to is irrelevant. If you want users of the table to have a precision of
T, you could protect users who edit more often than T by setting the timestamps
to some deterministically chosen point within T so that changes within that
interval are suppressed, but information would still leak about users who edit
less often. For T = 1 day, that's hardly acceptable, even on large wikis.

We could randomize the last few bits of the date, that would make it hard to
tell when it changes, but it would still be doable via statistical means...
really the only reliable way to mitigate this is to use a temporary table
instead of a view, and only update it once a day/week/whatever so that the
timing of the field changing becomes completely disconnected from the timing of
the edit.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
Wikibugs-l mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikibugs-l

Reply via email to