Hacker News new | past | comments | ask | show | jobs | submit login
Exercise to Detect Algorithmically Generated Domain Names (ipython.org)
44 points by boyd on July 16, 2014 | hide | past | favorite | 4 comments



In the output of line 85 there two punycoded[1] domain names: xn--g5t518j which is actually 微風 and xn--42cgk1gc8crdb1htg3d which is actually มอไซค์มือสอง.

As a result of the encoding, the domain names look junky while perhaps they aren't.

[1]: https://en.wikipedia.org/wiki/Punycode


มอไซค์มือสอง is, informally, second-hand motorcycles (or parts).


Awesome. Statistics are fun.

One little issue is that's assuming techniques like this aren't used to generate the domains, which is not only an obvious 'next step', but thanks to the 'weird' domains and the find-web-online-lol.info type results which look incredibly spammy, is probably already happening and has probably been happening for years (wouldn't surprise me with the amount of spam wielding Markov chain generators out there). That's definitely what I'd do if I was in the position of the VXer, if I was determined to use DNS at all.

Wonder if any of those are 'fast-flux'?


Really cool presentation.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: