quartz/content/notes/18-ML-in-IA-2.md

title	aliases	tags
18-ML-in-IA-2		lecture comp210

nefarious uses of ml

password guessing

• normally based on heuristics that are designed by humans
  • biased may not match true distributions of passwords
• leaked data can be used to "learn" what to guess
• gain insight into what users use as passwords

alternative - PassGan

• use statistical distribution of passwords then use this to generate guesses

new password generate, which also provides real world indicator of password st

bigger issues

where to from here