3 ms·
Good posted, upvoted. One clarification: > That's 25 users per hash Password choices are probably Zipf-distributed, so averages don't make a ton of sense.
by endtime 14y ago
Good posted, upvoted. One clarification:
> That's 25 users per hash
Password choices are probably Zipf-distributed, so averages don't make a ton of sense.
- dredmorbius 14y agoIt does if you're trying to estimate the size of the corpus based on the number of users. The arithmetic mean is specifically the value you'd want. n users times m users/password == total passwords (unduplicated) in the LinkedIn database. Zipf distribution would suggest that the pattern of reuse among passwords isn't normal, and that the median and mode are probably higher than the arithmetic mean.